PolarsCursor fails with the deprecated dtypes reader option #1111

Problem

PolarsCursor.execute(..., dtypes={...}) fails when the result is read as CSV.
The result set always passes its own schema_overrides (the column types from Athena's metadata) to polars.read_csv() / polars.scan_csv(), and Polars rejects receiving both the deprecated dtypes name and schema_overrides:

TypeError: `'read_csv'` received both `'dtypes'` and `'schema_overrides'` as arguments; `'dtypes'` was deprecated in version 0.20.31, use `'schema_overrides'` instead

PyAthena raises it as OperationalError after the query has run.
schema_overrides= works, so the only effect is on callers still using the alias deprecated since Polars 0.20.31 (PyAthena requires polars>=1.39.0).

The other code paths that special-case schema_overrides given to execute() (_get_timestamp_dtypes(), the new_columns handling in _read_csv() and the chunked CSV reader in pyathena/polars/result_set.py) do not recognize dtypes either.

Options

1. Treat dtypes as schema_overrides in the result set (drop the default schema_overrides and apply the same new_columns/timestamp rules).
2. Document schema_overrides as the supported name and leave the deprecated alias to Polars.

Found while reviewing #1102.