What problem does this PR solve?
When creating or updating datasets with custom embedding models (e.g.,
Ollama), the validation logic was too restrictive and prevented valid
models from being used. The previous implementation would reject valid
custom models if they weren't in the predefined list, even when they
existed in TenantLLMService.

Changes:
- Simplify and improve the embedding model validation flow in
create/update endpoints
- Check TenantLLMService for custom models before rejecting
- Make validation logic more consistent between create and update
operations

### What problem does this PR solve?

This fix allows users to successfully create and update datasets with
custom embedding models while maintaining proper validation checks.

### Type of change

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

---------

Co-authored-by: Jin Hai <haijin.chn@gmail.com>
Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>
Co-authored-by: liuhua <10215101452@stu.ecnu.edu.cn>