fix: Cast smaller integer types to int32/int64 on write for Spark compatibility

[somasays]

Summary

• Fixesattempting to write smallint/tinyint into int column results in incompatibility with other iceberg APIs #2791: Writing smaller integer types (uint8, int8, int16, uint16) to Iceberg IntegerType columns now correctly casts to int32/int64
• PyIceberg was preserving original Arrow types in Parquet files, causing Spark to fail with Unsupported logical type: UINT_8
• Added integer type widening logic in ArrowProjectionVisitor._cast_if_needed() following the same pattern as existing timestamp handling
• Only widening conversions are allowed (e.g., uint8 → int32, int32 → int64); narrowing conversions continue to be rejected via promote()

Test plan

• All 3041 unit tests pass
• Lint passes
• New parameterized test covers: uint8, int8, int16, uint16 → int32 and uint32, int32 → int64
• Existing test_projection_filter_add_column_demote still works (narrowing rejection)
• Manual verification: uint8 data written to IntegerType column produces int32 in Parquet file

Closes#2791

[somasays]

Verification with issue reproduction scenario

Source parquet type: uint8 Written parquet type: int32 ✓ (Spark compatible) Data integrity: preserved ✓ 

Tested with exact data from #2791 (uint8 column with values [None, None, None, None, 217, 163, 130, None, 69, 78]).