[parquet] reader perf - avoid some rle decode loop senarios in read_def_levels

can we skip rle decode loop when all values in a batch are valid (have max definition level)?

Start by locating the Parquet reader's read_def_levels entry point and inspect how the RLE decode loop handles batches whose values all have the maximum definition level. Confirm the expected behavior with the surrounding reader logic; done means those fully valid batches avoid unnecessary RLE decoding without changing definition-level results.

Written by the indexing model from the issue text.