Should ScanExec use Spark-compatible cast instead of DataFusion cast?

[andygrove]

What is the problem the feature request solves?

In ScanExec there can be a difference between the Parquet data type and the Spark data type being requested. In this case, we currently delegate to DataFusion's cast_with_options, but should we be using Comet's spark_cast method instead?

Describe the potential solution

No response

Additional context

No response