FEAT: LOB detection and fetch strategy for FetchAll and FetchMany

[gargsaumya]

Work Item / Issue Reference

AB#34162
AB#38110

GitHub Issue: #<ISSUE_NUMBER>

---

Summary

This pull request improves the handling of large object (LOB) columns when fetching data from SQL Server in the mssql_python/pybind/ddbc_bindings.cpp file. The main change is to detect LOB columns (such as large strings or binary data) and switch to a per-row fetching strategy using SQLGetData_wrap for those columns, ensuring correct streaming and memory usage. For non-LOB columns, the batch fetching logic remains unchanged, but now includes logic to handle LOBs appropriately during batch fetches.

LOB detection and fetch strategy:

• Added logic in both FetchMany_wrap and FetchAll_wrap to detect LOB columns by checking column data types and sizes, and to fall back to per-row fetch using SQLGetData_wrap when LOBs are present. This avoids buffer overflows and streams LOBs correctly. [1] [2]

• Updated calls to FetchBatchData in both wrappers to pass the list of detected LOB columns, so batch fetches can handle them appropriately. [1] [2]

Batch fetch improvements:

• Modified the signature of FetchBatchData to accept the LOB columns list and updated its logic to handle LOB columns differently: instead of throwing exceptions when buffer sizes are insufficient, it now streams LOB data using FetchLobColumnData. [1] [2] [3] [4] [5]

[sumitmsft]

Left few comments

[bewithgaurav]

lightly addressed some existing comments, requested one change