Please write user-readable short description of the changes:

• Problem
  • Python UDF/UDA code that returns numpy scalar ints (e.g. numpy.int64) fails at the Python→Proton boundary with “Failed to convert Python object to long”.
• What changed
  • src/CPython/ConvertDatatypes.cpp:336:
    • Keep fast-path for PyLong.
    • For non-PyLong integer-like objects, coerce via PyNumber_Index() and then convert, so numpy scalars (and other index providers) work.
    • Improve error reporting by surfacing Python exception text when coercion fails.
  • src/CPython/tests/gtest_data_convert.cpp:163:
    • Add regression tests for inserting numpy.int64(15) into int64 and numpy.uint64(15) into uint64.