Use ColumnBase.create in from_arrow, follow-ups in pyarrow 19 decimal support

[mroeschke]

Description

• Use ColumnBase.create in ColumnBase.to_arrow to more reliably preserve dtype metadata
• Follows up with Require pyarrow 19 or greater #21183 to eliminate more decimal32/64 special casing
  • Fixed a bug where ArrowDtype.to_arrow was still always returning a pa.Decimal128 object

The added tests to conftest-patch.py use "object" in the test body as pandas does not support a native decimal type. This is incongruent as we interpret object as string leading to incorrect results, which I think is becoming a defining different when using object in cuDF vs pandas

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[vyasr]

/merge