Add /llm POST Endpoint

[CelestineAkpanoko]

Description:
Implements a new /llm POST endpoint for generic LLM inference using Groq Cloud (LLaMA 3.1). Separates request and response models into src/core/api/models/llm/request.py and src/core/api/models/llm/response.py. Tests are organized in tests/llm/unit/, tests/llm/integration/, and tests/llm/e2e/. Fixed test discovery and async mocking issues.

Changes:

• Added LLMRequest in src/core/api/models/llm/request.py.
• Added LLMResponse in src/core/api/models/llm/response.py.
• Added src/core/api/routes/llm.py with new import paths.
• Updated src/main.py to include the LLM router.
• Added unit tests in tests/llm/unit/test_llm_models.py.
• Added integration tests in tests/llm/integration/test_llm_routes.py.
• Added E2E tests in tests/llm/e2e/test_llm_workflow.py.
• Added tests/llm/conftest.py for fixtures.
• Added pytest.ini to configure pytest-asyncio, set pythonpath = ., and add testpaths = tests/llm and e2e marker.
• Added docs/llm.md
• Fixed failing tests by correcting endpoint to /llm/ and using AsyncMock.

Testing:

• Unit tests for LLMRequest and LLMResponse models.
• Integration tests for /llm endpoint with mocked Groq responses.
• E2E tests for full workflow (prompt + context).
• All tests pass (pytest tests/llm/).

Checklist:

• Tests written and passing
• Documentation updated
• Code follows project style guidelines
• External services (Groq) mocked

[LokiMetaSmith]

This pull needs to be more generic.
It needs to support more than just groq and it shouldn't add in groq as a dependency.

For example, there is no reason why this couldn't run on ollama, llama.cpp, open.ai, or any of the vast number of other AI providers.

good start though.