perf(parquet): batch DELTA_BYTE_ARRAY encoding

[fallintoplace]

Summary

• Batch DELTA_BYTE_ARRAY prefix lengths and suffixes in chunks of 256 values.
• Reuse fixed scratch space for DELTA_LENGTH_BYTE_ARRAY lengths.
• Preserve the previous value across batch boundaries and separate Put calls.
• Add a boundary round-trip test and representative benchmarks.

Why

DELTA_BYTE_ARRAY was calling the prefix and suffix encoders once per value. Large pages therefore paid for many small calls and temporary length slices.

Apache Arrow C++ already uses fixed-size batching for this encoder. This change follows the same shape while keeping the existing Go encoding format and last-value behavior.

Correctness

• The encoded layout is unchanged: prefix lengths first, followed by the suffix stream.
• The first value still has a zero prefix.
• Empty suffixes still work as before.
• The new test compares one large Put call with Put calls split around a batch boundary, then decodes and compares every value.

Benchmark

64K values on an Apple M1 Pro, 1 second per sample, 3 samples:

Case	Before	After	Change
prefix-heavy	2.09 ms/op	1.75 ms/op	~16% faster
low-prefix	1.68 ms/op	1.47 ms/op	~12% faster

Allocation counts stayed the same in both cases.

Checks

• PARQUET_TEST_DATA=<parquet-testing-data> go test ./parquet/...
• go test -race ./parquet/internal/encoding -count=1
• git diff --check

No public API changes.