Parquet dictionary fallback heuristics #9699

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
The dictionary encoding can be inefficient for Parquet column data with high cardinality.
To mitigate this, parquet-java implements fallback based on the evaluation of the size of plainly encoded data for the first data page versus its dictionary-based encoding: if compression is not achieved, the column encoder falls back to the plain encoding.

Describe the solution you'd like
Add an option to enable dictionary fallback decided on compression efficiency, with a possibility to extend it with other behaviors in the future.

Describe alternatives you've considered
Implement the heuristic without an option. Since the added computation has some cost and is not guaranteed to bring benefits (in fact, it may be detrimental over the entire page chunk), this may be undesirable for some applications. The new encoding behavior is also a substantial change in output, so it's better introduced through an option.