MINOR: Reuse RLE decoder buffers for packed runs

[divjotarora]

Rationale for this change

RunLengthBitPackingHybridDecoder allocated integer and byte buffers for every packed run. This
decoder is used for repetition and definition levels and dictionary IDs, so repeated packed runs
created avoidable allocation and GC pressure.

What changes are included in this PR?

The decoder now grows and reuses its packed-run buffers, tracks the active value position
independently of buffer capacity, and reads directly from the input stream. It preserves existing
short-input zero padding and EOF behavior. A focused JMH benchmark and regression tests cover
smaller runs following larger runs and truncated final groups.

Are these changes tested?

Yes. The existing parquet-column tests pass. The new benchmark shows 27-35% speedup and reduced allocations:

> ./parquet-benchmarks/run.sh all \
    org.apache.parquet.benchmarks.RunLengthBitPackingHybridDecodingBenchmark \
    -wi 3 -i 5 -f 2 -prof gc

   Bit width                       Before ops/s                        After ops/s    Change    Before B/value    After B/value
  ━━━━━━━━━━━  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━  ━━━━━━━━  ━━━━━━━━━━━━━━━━  ━━━━━━━━━━━━━━━
           1    125,889,985.415 ± 1,443,739.608      173,074,464.937 ± 470,051.407    +37.5%    13.501 ± 0.001    0.002 ± 0.001
  ───────────  ─────────────────────────────────  ─────────────────────────────────  ────────  ────────────────  ───────────────
           8      130,416,053.717 ± 754,362.833      175,155,689.891 ± 499,245.748    +34.3%    13.751 ± 0.001    0.002 ± 0.001
  ───────────  ─────────────────────────────────  ─────────────────────────────────  ────────  ────────────────  ───────────────
          16    132,141,608.282 ± 5,365,801.318    167,817,248.033 ± 1,318,258.836    +27.0%    13.751 ± 0.797    0.002 ± 0.001

Are there any user-facing changes?

No. This is an internal performance improvement with no API or format changes.