Add n5_label_multiset codec for N5 legacy label multiset payload

[mkitti]

Note

This is a draft PR pending review of #54 (n5_varlen). The n5_label_multiset codec is designed to be used as the inner array-to-bytes codec inside n5_varlen and depends on that codec being accepted first.

Summary

Registers the n5_label_multiset codec, which serializes label_multiset arrays using the N5 legacy binary format produced by imglib2-label-multisets (LabelUtils.serializeLabelMultisetTypes()).

Codec chain

n5_label_multiset is not used standalone. It lives inside n5_varlen's inner codec pipeline:

{
  "name": "n5_varlen",
  "configuration": {
    "codecs": [
      {"name": "n5_label_multiset"},
      {"name": "gzip", "configuration": {"level": 6}}
    ]
  }
}

n5_varlen handles the N5 block header; n5_label_multiset handles the payload serialization.

N5 legacy payload format

argMaxSize             (int32 BE, always 0)
argMax[argMaxSize]     (int64 BE each — legacy; skipped on read)
listEntryOffsets[N]    (int32 BE each)
listData               (all little-endian)

Note on byte-order asymmetry: argMaxSize, argMax, and listEntryOffsets are big-endian; listData is little-endian. This is a historical artifact of the original Java implementation.

Related PRs

• Add n5_varlen codec for N5 varlength block format #54 — n5_varlen codec (prerequisite; please review first)
• Add label_multiset data type and Zarr-native codec #55 — label_multiset data type and Zarr-native codec

Test plan

• Review Add n5_varlen codec for N5 varlength block format #54 (n5_varlen) first
• Confirm codecs/n5_label_multiset/schema.json validates correctly
• Verify annotated binary layout against a real N5 varlength block from imglib2-label-multisets

🤖 Generated with Claude Code

[jbms]

You also need to define a zarr data type for this.

[mkitti]

The data type is in #55