- Unified configuration format: All services now use the same image
configuration structure for consistency.

- Private registry support: Added imagePullSecrets to enable pulling
images from private registries.

- Per-service flexibility: Each service can override image-related
parameters independently.

### What problem does this PR solve?

_Briefly describe what this PR aims to solve. Include background context
that will help reviewers understand the purpose of the PR._

### Type of change

- [ ] Bug Fix (non-breaking change which fixes an issue)
- [ ] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [ ] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):

### What problem does this PR solve?

Adds configuration options to the RAGFlow Helm chart to set the Redis
data retention policies. By default this feature is disabled to maintain
support with older Kubernetes versions.

https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/#persistentvolumeclaim-retention

### Type of change

- [X] New Feature (non-breaking change which adds functionality)

### What problem does this PR solve?

Previous version created labels which were dependent on the specific
Helm chart version such as:
```
volumeClaimTemplates:
- metadata:
    name: redis-data
    labels:
      helm.sh/chart: ragflow-0.2.3-dev.0.opensearch-test.4
      app.kubernetes.io/name: ragflow
      app.kubernetes.io/instance: test-1
      app.kubernetes.io/version: "9a04408"
      app.kubernetes.io/managed-by: Helm
      app.kubernetes.io/component: redis
```
which causes `helm upgrade` commands to fail with
```
Upgrade "test-1" failed: cannot patch "test-1-ragflow-redis" with
kind StatefulSet: StatefulSet.apps "test-1-ragflow-redis" is
invalid: spec: Forbidden: updates to statefulset spec for fields
other than 'replicas', 'ordinals', 'template', 'updateStrategy',
'persistentVolumeClaimRetentionPolicy' and 'minReadySeconds' are
forbidden
```
because the labels changed on upgrade.

This fix uses a reduced set of labels to prevent upgrade failures.

### Type of change

- [X] Bug Fix (non-breaking change which fixes an issue)

### What problem does this PR solve?

Fix Kubernetes liveness probe on the OpenSearch container. The previous
HTTP probe received an 401 response from the OpenSearch API which
treated as a failure and caused the container to be restarted every 20
minutes.

### Type of change

- [X] Bug Fix (non-breaking change which fixes an issue)

### What problem does this PR solve?

Switch to Kubernetes StatefulSet resources for MySQL, Minio and vector
DB since these are stateful application components. This makes
operations such as helm upgrade smoother since the default container
update strategy becomes a sequential rolling update of each pod.

Also fixes a bug in the name template for the Minio stateful set
resource to align it with the naming convention used for other
components.

### Type of change

- [X] Bug Fix (non-breaking change which fixes an issue)

### What problem does this PR solve?

Adds OpenSearch support to the RAGFlow Helm chart based on
https://github.com/infiniflow/ragflow/pull/7140 and the existing
Elasticsearch support in the Helm chart.

### Type of change
- [X] New Feature (non-breaking change which adds functionality)

### What problem does this PR solve?

### Type of change

- [X] New Feature (non-breaking change which adds functionality)

Modified the chart to retain persistent volumes by default when the
chart is uninstalled, following established best practices in the Helm
community (e.g., Bitnami charts)

### What problem does this PR solve?

Previously, deleting the helm chart would automatically remove all
persistent data, which poses a risk of accidental data loss.

### Rationale

This change aligns with industry standards to safeguard data by
requiring explicit action to remove persistence, rather than making
deletion the default behavior.

### Impact: 

Users who intentionally want to remove persistent data will need to do
so manually or by setting appropriate flags during chart uninstallation.

### Type of change

- [ ] Bug Fix (non-breaking change which fixes an issue)
- [x] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [ ] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):

### What problem does this PR solve?

Fixes bug & regression introduced by [PR #7187 - refactor: Update Redis
configuration to use StatefulSet instead of deployment with
pvc](https://github.com/infiniflow/ragflow/pull/7187):

1. Fixes bug #7403 - `redis.persistence.enabled` missing from
`helm/values.yaml` causes helm error:

[ERROR] templates/: template: ragflow/templates/redis.yaml:55:24:
executing "ragflow/templates/redis.yaml" at
<.Values.redis.persistence.enabled>: nil pointer evaluating interface
{}.enabled

2. Fixes regression: reverts hardcoded redis.storage.capacity value back
to using variable `redis.storage.capacity` from `helm/values.yaml`.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
- [ ] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [ ] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):

### What problem does this PR solve?

This PR changes Redis to be a statefulset. In some situation when we
Redis pod gets rescheduled to another Node, it gets stuck in pending
state due to the PVC attached to another Kubernetes node.

### Type of change

- [ ] Bug Fix (non-breaking change which fixes an issue)
- [ ] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [X] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):

### What problem does this PR solve?

Fix Helm Ingress template; Trying to access a global variable within a
loop
Fix #6191

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
- [ ] New Feature (non-breaking change which adds functionality)
- [ ] Documentation Update
- [ ] Refactoring
- [ ] Performance Improvement
- [ ] Other (please describe):

fix redis pvc in helm deployment

### What problem does this PR solve?

_Briefly describe what this PR aims to solve. Include background context
that will help reviewers understand the purpose of the PR._

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

### What problem does this PR solve?

Adds a new Kubernetes Service resource to the Helm chart which
specifically targets the RAGFlow API. This feature useful for cases
where you want to expose the RAGFlow HTTP API separately from the web
interface, for example if RAGFlow is running behind an authenticating
proxy it allows a route to bypass the proxy (e.g. by defining a separate
ingress resource which forwards to the separate API-only k8s service
added here) to provide RAGFlow API access. This is still secure since
API access is already authenticated by API keys inside RAGFlow itself.

### Type of change

- [X] New Feature (non-breaking change which adds functionality)

### What problem does this PR solve?

Changed elasticsearch image url to speed up image downloading. 

### Type of change

- [x] Refactoring

### What problem does this PR solve?

Address #4391 

### Type of change

- [x] Other (please describe):

### What problem does this PR solve?

The initial Helm chart implementation added in #3815 suffers from an
issue where the 5GB data volume for the SQL DB is filled up with
[binlog](https://dev.mysql.com/doc/refman/8.4/en/binary-log.html) files
after just a few days. Since the app uses a non-replicated SQL DB config
I think it makes sense to disable the binlog in the SQL DB container.
This is achieved by simply adding the required argument to the container
startup command.

### Type of change

- [X] Bug Fix (non-breaking change which fixes an issue)

### What problem does this PR solve?

Add's a Helm chart for deploying RAGFlow on Kubernetes. 

Closes #864. 

### Type of change

- [X] New Feature (non-breaking change which adds functionality)