fix unpickle_bytes_columns crash on variable-size binary columns

[TeunHuijben]

map_elements without return_dtype infers the column type from the first deserialized element. When a binary column contains numpy arrays of different shapes (e.g. cell masks after a GEFF roundtrip), polars locks in the type from the first array and raises SchemaError on any subsequent array with a different shape.

The function already had a fallback loop to handle variable-size objects — it was simply unreachable because map_elements crashed before returning.

Fix: add return_dtype=pl.Object to skip type inference. The existing fallback loop then re-casts uniform columns to typed polars Series and silently keeps variable-size columns as pl.Object.

Regression test: test_unpickle_bytes_columns_variable_size_arrays — two differently-shaped bool arrays in a binary column, previously raised SchemaError.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.33%. Comparing base (47f450a) to head (3348d7b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #273      +/-   ##
==========================================
+ Coverage   87.29%   87.33%   +0.04%     
==========================================
  Files          56       56              
  Lines        4713     4713              
  Branches      832      832              
==========================================
+ Hits         4114     4116       +2     
+ Misses        380      378       -2     
  Partials      219      219              

☔ View full report in Codecov by Sentry.
📢 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.