fix: @tool_method on pydantic base models

[monoxgas]

Notes

• Fix support for applying @tool_method to pydantic base model classes.
• Migrate a bunch of optional import logic to be lazy.
• Clean up dependencies and typing for optional packages.

---

Generated Summary

• Refactored type annotations and type ignore comments across multiple modules (e.g. using string annotations for forward references and adjusting spacing in type ignore directives).
• Updated dependency extras in pyproject.toml: moved elasticsearch and pandas from the default dependency group to the "rigging[data]" and "rigging[llm]" extras.
• Modified function signatures to import dependencies locally with appropriate error messages when dependencies (like pandas or elasticsearch) are missing.
• Revised the ToolMethod implementation in rigging/tools/base.py to correctly handle bound and unbound tool methods, using a DEFAULT_CATCH_EXCEPTIONS set.
• Added or adjusted default parameters (e.g. default log_level in logging configuration) to improve API consistency.
• Improved tests in tests/test_tool.py to validate tool method binding and decorator argument handling.

This summary was generated with ❤️ by rigging