Fix filtering on parquet file containing a struct column

[rjzamora]

Dask does not support filtering on a PyArrow struct column, but it should support filtering on other columns when a struct column is present. This PR fixes a bug in the way we align field names with field indices when struct columns are present.

• Closes Pyarrow filters not working on dataframes with pa.struct() dtypes #11652
• Tests added / passed
• Passes pre-commit run --all-files

[github-actions]

Unit Test Results

See test report for an extended history of previous test failures. This is useful for diagnosing flaky tests.

     15 files  ± 0       15 suites  ±0   4h 28m 29s ⏱️ - 1m 25s
 17 156 tests + 1   15 961 ✅ + 1   1 195 💤 ±0  0 ❌ ±0 
211 366 runs  +12  194 191 ✅ +15  17 175 💤  - 3  0 ❌ ±0 

Results for commit f09cc81. ± Comparison against base commit 1e6ed84.

[phofl]

Thanks!

I did not know that we aren't supporting filtering on struct columns, do you know if we pay attention not to push filters into read_parquet if they are applied to a struct column?

[rjzamora]

I did not know that we aren't supporting filtering on struct columns

Yeah, I honestly hadn't thought much about filtering on a struct column until today :/
My statement is mostly an observation that there is currently no way to write a DNF filter argument that would be translated into the appropriate PyArrow filter expression (at least I don't think there is), and we would handle such a filter wrong in _get_rg_statistics anyway... All the more reason for a direct from_arrow_dataset API ;)