FEAT: adding perf improvements

[gargsaumya]

Work Item / Issue Reference

AB#<WORK_ITEM_ID>

GitHub Issue: #<ISSUE_NUMBER>

---

Summary

This pull request introduces several performance and maintainability improvements to the MSSQL Python driver, focusing on optimizing row creation, reducing repeated lookups of Python types and settings, and ensuring proper cursor configuration. The changes span both the Python and C++ layers, with new caching mechanisms and more efficient resource usage.

Python-side row creation optimizations

• Added caching for column name maps, converter maps, and settings snapshots in the Cursor class to avoid repeated computation during row fetching. These caches are now built and reused for all fetch operations (fetchone, fetchmany, fetchall). [1] [2]
• Updated row creation logic to use these cached optimizations, resulting in more efficient instantiation of Row objects. [1] [2] [3]

C++/pybind11 Python type caching

• Introduced a PythonObjectCache namespace in ddbc_bindings.cpp to cache references to Python classes (datetime, date, time, decimal, uuid). This reduces overhead and avoids issues during Python finalization. All type checks and conversions now use the cached objects. [1] [2] [3] [4] [5] [6] [7]

Cursor configuration and resource management

• Changed statement execution in C++ to use forward-only cursors instead of static cursors, improving compatibility and performance for result set traversal. [1] [2]
• Ensured __exit__ in the Python Cursor class consistently returns None for proper context manager behavior.

Miscellaneous improvements

• Improved error handling and result set initialization in the Python execute method, ensuring column metadata is properly described and available for optimizations.
• Cached the decimal separator in the C++ data retrieval logic to minimize repeated system calls when parsing numeric data.

[gargsaumya]

This is not needed anymore, perf PR is merged. Closing this draft.