fix: use only supported operators in metadata filter system prompts (#19195)

api/core/rag/retrieval/template_prompts.py

     ### Job Description',
     You are a text metadata extract engine that extract text's metadata based on user input and set the metadata value
     ### Task
-    Your task is to ONLY extract the metadatas that exist in the input text from the provided metadata list and Use the following operators ["=", "!=", ">", "<", ">=", "<="] to express logical relationships, then return result in JSON format with the key "metadata_fields" and value "metadata_field_value" and comparison operator "comparison_operator".
+    Your task is to ONLY extract the metadatas that exist in the input text from the provided metadata list and Use the following operators ["contains", "not contains", "start with", "end with", "is", "is not", "empty", "not empty", "=", "≠", ">", "<", "≥", "≤", "before", "after"] to express logical relationships, then return result in JSON format with the key "metadata_fields" and value "metadata_field_value" and comparison operator "comparison_operator".
     ### Format
     The input text is in the variable input_text. Metadata are specified as a list in the variable metadata_fields.
     ### Constraint

api/core/workflow/nodes/knowledge_retrieval/knowledge_retrieval_node.py

     METADATA_FILTER_COMPLETION_PROMPT,
     METADATA_FILTER_SYSTEM_PROMPT,
     METADATA_FILTER_USER_PROMPT_1,
+    METADATA_FILTER_USER_PROMPT_2,
     METADATA_FILTER_USER_PROMPT_3,
 )
 from core.workflow.nodes.llm.entities import LLMNodeChatModelMessage, LLMNodeCompletionModelPromptTemplate
 from core.workflow.nodes.llm.node import LLMNode
-from core.workflow.nodes.question_classifier.template_prompts import QUESTION_CLASSIFIER_USER_PROMPT_2
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from libs.json_in_md_parser import parse_and_check_json_markdown
             )
             prompt_messages.append(assistant_prompt_message_1)
             user_prompt_message_2 = LLMNodeChatModelMessage(
-                role=PromptMessageRole.USER, text=QUESTION_CLASSIFIER_USER_PROMPT_2
+                role=PromptMessageRole.USER, text=METADATA_FILTER_USER_PROMPT_2
             )
             prompt_messages.append(user_prompt_message_2)
             assistant_prompt_message_2 = LLMNodeChatModelMessage(

api/core/workflow/nodes/knowledge_retrieval/template_prompts.py

     ### Job Description',
     You are a text metadata extract engine that extract text's metadata based on user input and set the metadata value
     ### Task
-    Your task is to ONLY extract the metadatas that exist in the input text from the provided metadata list and Use the following operators ["=", "!=", ">", "<", ">=", "<="] to express logical relationships, then return result in JSON format with the key "metadata_fields" and value "metadata_field_value" and comparison operator "comparison_operator".
+    Your task is to ONLY extract the metadatas that exist in the input text from the provided metadata list and Use the following operators ["contains", "not contains", "start with", "end with", "is", "is not", "empty", "not empty", "=", "≠", ">", "<", "≥", "≤", "before", "after"] to express logical relationships, then return result in JSON format with the key "metadata_fields" and value "metadata_field_value" and comparison operator "comparison_operator".
     ### Format
     The input text is in the variable input_text. Metadata are specified as a list in the variable metadata_fields.
     ### Constraint