fix: improve error logging for chat stream failures

[beanrepo]

This PR fixes a streaming error logging issue in the LLM chat flow. Errors raised while consuming streamed chat responses were being propagated to the app layer and shown in the frontend error box, but they were not always emitted in the Python logs.

Technical Details

The root cause was that CloudLLM.chat_stream() returned the internal generator directly, so exceptions raised during generator iteration happened outside its try/except block.

Changes:

• Updated src/arduino/app_bricks/cloud_llm/cloud_llm.py

  • CloudLLM.chat_stream() now uses yield from so stream iteration is covered by the logging/error handling block.
  • Added _handle_stream_error() as the common stream error handling hook.

• Updated src/arduino/app_bricks/llm/local_llm.py

  • LargeLanguageModel.chat_stream() now delegates to super().chat_stream(...), sharing the same streaming/logging path as CloudLLM.
  • Kept local LLM-specific behavior isolated: filtering <think> chunks and handling OpenAI-compatible API errors via _handle_api_error().

• Added brick-scoped regression tests:

  • tests/arduino/app_bricks/cloud_llm/test_cloud_llm_stream_logging.py
  • tests/arduino/app_bricks/llm/test_local_llm_stream_logging.py

Validation

• python3 -m pytest tests/arduino/app_bricks/cloud_llm/test_cloud_llm_stream_logging.py tests/arduino/app_bricks/llm/test_local_llm_stream_logging.py
• python3 -m ruff check src/arduino/app_bricks/cloud_llm/cloud_llm.py src/arduino/app_bricks/llm/local_llm.py tests/arduino/app_bricks/cloud_llm/test_cloud_llm_stream_logging.py tests/arduino/app_bricks/llm/test_local_llm_stream_logging.py

[lucaronca]

Approved with an optional comment 👍🏻