Enhance decoders to support value-skipping

When the parquet scanner can use the page index it is sometimes needed to skip values from the pages. We can just drop these values since we'll never need them.

The decoders don't support value-skipping at the moment, therefore we have to decode each and every value even if we know that we will never need many of them.

The decoders should have methods for value-skipping to make scanning more efficient.