Preserve field metadata when coercing INT96 timestamps

[ryzhyk]

Which issue does this PR close?

• Closes Int96Coercer drops the metadata of struct, list and map fields. #24786.

Rationale for this change

Int96Coercer rebuilds every struct, list and map field with Field::new_struct, Field::new_list and Field::new, which start from empty metadata. Leaf fields go through field_with_new_type, which clones the field, so only container fields are affected, and only when the file holds an INT96 column: the same file read without coerce_int96 keeps its metadata.

What changes are included in this PR?

Carry current_field.metadata() across in the three constructors.

Are these changes tested?

yes

Are there any user-facing changes?

[codecov-commenter]

Codecov Report

❌ Patch coverage is 96.96970% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 81.53%. Comparing base (4ada0dc) to head (8bb2406).

Files with missing lines	Patch %	Lines
...tafusion/datasource-parquet/src/schema_coercion.rs	96.96%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #24790      +/-   ##
==========================================
- Coverage   81.53%   81.53%   -0.01%     
==========================================
  Files        1123     1123              
  Lines      406041   406074      +33     
  Branches   406041   406074      +33     
==========================================
+ Hits       331049   331073      +24     
- Misses      55631    55636       +5     
- Partials    19361    19365       +4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.