Comet should choose the most suitable Parquet scan implementation automatically

[andygrove]

What is the problem the feature request solves?

We currently support three different Parquet scan implementations:

• native_comet
• native_datafusion
• native_iceberg_compat

We really shouldn't expose the details of this or expect users to choose which implementation to use, since this is a very low level technical detail. It would be better if Comet just used the most appropriate one based on the Parquet schema and knowledge of supported features and known issues in the different scan implementations. For example, some of these scans do not support S3 or encryption currently.

Describe the potential solution

No response

Additional context

No response