Refactor speech_to_text.py to implement a singleton ASR pipeline, enhance error handling, and introduce SpeechToTextTool for better integration. Update spreadsheet_tool.py to support querying and improve parsing functionality, including CSV support. Enhance video_processing_tool.py with new tasks for metadata extraction and frame extraction, while improving object detection capabilities and initialization checks.

add smolagents docstrings

refactor: improve ASR pipeline initialization and transcription handling

feat: add speech-to-text transcription functionality using Hugging Face ASR pipeline