fix: suppress pyarrow nullability warning for primary keys

[Sam-Razavi]

Description

Fixes #3581.

When remove_nullability_adapter is used with the PyArrow backend, dlt correctly keeps reflected primary-key columns as nullable=False. PyArrow does not retain primary-key metadata and reports those fields as nullable, which causes a misleading Arrow-schema conflict warning.

This PR suppresses only that expected mismatch when:

• the computed dlt column is a primary key
• dlt has nullable=False
• Arrow has nullable=True

Schema merging remains unchanged, and primary-key columns remain non-nullable.

Related Issues

• Fixes Using remove_nullability_adapter with PyArrow triggers warning on tables with primary key #3581

Additional Context

Added a SQL + PyArrow regression test using remove_nullability_adapter that verifies the reflected id column remains a non-nullable primary key and that no Arrow-schema merge message is emitted.

Local checks completed:

• Focused SQL + PyArrow regression test passed
• Existing test_remove_nullability passed
• flake8 for changed files passed
• git diff --check passed

Targeted mypy did not complete locally within five minutes and produced no diagnostics, so CI should verify that check.