Optimize writing RLE runs in parquet column descriptors

[raunaqmorarka]

Description

Optimize writing RLE runs in parquet column descriptors

Use information about nullability of Blocks to write RLE runs
for repetition and definition levels more efficiently in parquet writer

BenchmarkParquetFormat#write UNCOMPRESSED
                      Before                          After
LINEITEM                  293.0MB/s ± 2869.6kB/s (0.96%)  312.9MB/s ± 2869.2kB/s (0.90%) (N = 10, α = 99.9%)
MAP_VARCHAR_DOUBLE        345.4MB/s ± 3275.7kB/s (0.93%)  359.6MB/s ± 5555.4kB/s (1.51%) (N = 10, α = 99.9%)
LARGE_MAP_VARCHAR_DOUBLE  402.0MB/s ± 6815.6kB/s (1.66%)  448.6MB/s ± 4808.3kB/s (1.05%) (N = 10, α = 99.9%)
MAP_INT_DOUBLE            606.2MB/s ± 2136.1kB/s (0.34%)  676.1MB/s ± 5620.8kB/s (0.81%) (N = 10, α = 99.9%)
LARGE_ARRAY_VARCHAR       257.8MB/s ± 9303.4kB/s (3.52%)  275.7MB/s ± 2583.1kB/s (0.91%) (N = 10, α = 99.9%)

Additional context and related issues

Release notes

( ) This is not user-visible or is docs only, and no release notes are required.
( ) Release notes are required. Please propose a release note for me.
(x) Release notes are required, with the following suggested text:

# Hive, Hudi, Delta Lake, Iceberg
* Improve performance of writing parquet files. ({issue}`22089`)

[raunaqmorarka]

~4.2% CPU time improvement in insert benchmarks