Adding support for typed dict in schema conversion

[lucyliulee]

What does this PR do?

Adds TypedDict support to the schema converter, enabling proper JSON schema generation for tools with nested object parameters. This fixes a critical issue where OpenAI models reject list[dict] parameters because the generated schema lacks property definitions required by OpenAI.

Problem:

• Using list[dict] in tool parameters generates {"type": "object"} without properties
• OpenAI's API rejects this with: additionalProperties is required to be supplied and to be false
• This caused tools to fail with GPT models while working with Gemini

Solution:

• Add TypedDict detection and schema generation in schema_converter.py
• Generate proper schemas with properties, required, and additionalProperties: false
• Add warnings when developers use dict or list[dict] types (pointing to TypedDict as the fix)
• Propagate strict flag through recursive type resolution

Example:

  # Before (broken with OpenAI)                                                                                                                                              
  items: Annotated[list[dict], "Items to order"]                                                                                                                             
  # Generated: {"type": "array", "items": {"type": "object"}}  ❌                                                                                                            
                                                                                                                                                                             
  # After (works with all models)                                                                                                                                            
  class MenuItem(TypedDict):                                                                                                                                                 
      menu_item_id: str                                                                                                                                                      
      quantity: int                                                                                                                                                          
                                                                                                                                                                             
  items: Annotated[list[MenuItem], "Items to order"]                                                                                                                         
  # Generated: {"type": "array", "items": {"type": "object", "properties": {...}, "additionalProperties": false}} 

Type of change

• Bug fix
• New feature
• Breaking change
• Documentation
• Other: ___________

Testing

Unit Tests (13 new tests):
uv run pytest tests/test_llm_agent_schema_converter.py -v

• TestIsTypedDict - TypedDict detection (4 tests)
• TestTypedDictSchema - Schema generation for simple, nested, optional TypedDicts (5 tests)
• TestDictWarnings - Warnings for dict and list[dict] usage (2 tests)
• TestFunctionToolWithTypedDict - End-to-end tool schema generation (2 tests)

Integration Tests:
uv run python line/llm_agent/scripts/test_provider.py --tests nested_objects,nested_objects_fail

• nested_objects - Validates TypedDict works with GPT 5.2, Claude, Gemini 2.5/3
• nested_objects_fail - Validates list[dict] correctly fails with OpenAI (negative test)

Checklist

• I have read the contributing guidelines
• I have added tests that prove my fix is effective or that my feature works
• I have formatted my code with make format