raw2features 0.2.0

This release extends named-channel multiplex support with a channelwise strategy, adds
several pinned patch and slide encoders, and includes a set of correctness and
release-safety fixes. The v0.1 store layout and existing brightfield workflows remain
unchanged.

Correctness and reproducibility

• Redact command-line secrets and persist credential-free canonical source URIs while
  still supporting query-authenticated HTTP Zarr stores in memory.
• Bind stores and receipts to their source and requested output, reject duplicate cohort
  output IDs before dispatch, make receipt writes atomic, and propagate worker-startup
  failures to the command exit status.
• Enforce immutable model revisions at load time. Patch and slide outputs now carry a
  per-model fingerprint covering their effective weights, preprocessing, pooling,
  dimension, resolved AMP, and loader contract; a stale or legacy-unfingerprinted output
  is recomputed without changing grid_hash or unrelated model arrays.
• Verify composite checkpoints before deserialisation, validate expected embedding
  dimensions during completion checks, and mark SEAL experimental until its upstream
  base-model fetch can be pinned end to end.
• Make colliding human-readable grid labels unique, produce explicitly requested
  thumbnails/QC/GeoJSON/slide embeddings when missing, and validate receipts against the
  requested grid rather than unioning same-named models across grids.
• Harden partial-store recovery and conformance validation, preserve per-axis scale in
  slide-relative STAMP coordinates, place clipped border pixels at their true offsets,
  reject invalid AMP/batch/geometry settings before model loading, and avoid attributing
  an enclosing unrelated Git checkout to an installed wheel.

Models and interoperability

• Add a pluggable multiplex-strategy layer and a built-in channelwise strategy for
  applying registered RGB patch encoders to named-channel multiplex tissue images. It
  records positional channel identity (including a strict metadata-file fallback), marker
  selection and order, normalization, RGB conversion, base-model identity, mean/concat
  aggregation, and model-agnostic slide pooling without changing brightfield outputs.
• Bind native multiplex outputs to the complete effective positional panel. Nuclear
  masking accepts one unambiguous nuclear channel or a recognized same-stain group, such
  as DNA1/DNA2 or repeated DAPI channels. It averages each multi-channel group in float32
  and binds every physical index into grid identity; --no-seg remains supported.
• Add pinned, forward-validated H0-mini, KEEP, OpenMidnight, and OpenPath patch encoders,
  with their licenses, access requirements, preprocessing, and weight hashes recorded.
• Add the paired GigaPath-Flash tile and slide encoders with immutable weight and loader
  pins, and align original GigaPath's default extraction with its released whole-slide
  path: non-overlapping 256-pixel tiles on a 256-pixel lattice, followed by a centre crop
  to 224. LongNet receives the stored level-0 coordinates and uses each tile's level-0
  extent as its runtime positional divisor, normalizing raw2features' coordinate units
  into the tile-index frame of upstream's default 256-pixel tiling.
• Repair HIPT downloads by replacing the unavailable unofficial mirror with Mahmood Lab's
  commit-pinned, SHA-verified Git LFS checkpoint and record its Commons Clause accurately.
• Correct cached OME-Zarr reads for noncanonical spatial axis order. The reader accepts
  x/y in either order (with an optional channel axis) and warns whenever another
  non-singleton axis is reduced to index zero. Unnamed omero.channels entries now
  retain their physical positions instead of shifting later marker identities.
• Refresh the development lock to current ngff-zarr/ome-zarr releases while retaining
  the existing declared compatibility floor.
• Choose the float-pixel [0,1] versus [0,255] convention once per opened slide instead
  of independently for every patch.

API, documentation, and release safety

• Bless embed_slide as the high-level Python entry point while retaining run_slide as
  the explicit single-grid primitive, resolve injected embedders from their own model
  specifications, and make inline -s/--slide-encoder requests produce missing outputs
  on otherwise-complete runs.
• Correct the multi-grid, additive-append, readback, and SLURM documentation, and use the
  compact PNG project banner so GitHub and PyPI render it consistently.
• Add a genuine dependency-lean import/CLI check, built-wheel and source-distribution
  smoke installs, package metadata validation, and a release tag/version gate.

Compatibility and migration

• Existing v0.1 stores remain readable; the store layout/schema and model-independent
  grid-identity contract are unchanged. Multiplex nuclear masking now binds the resolved
  physical nuclear-channel indices. Averaging a recognized same-stain group is an
  intentional value change from v0.1 and therefore creates a new grid. Compatible
  historical native-multiplex grids remain appendable only when the original OME panel
  proves that the v0.1 selector used the same single physical channel.
• Original GigaPath now defaults to the released 256-pixel non-overlapping extraction
  lattice followed by a centre crop to 224. Existing 224-pixel grids remain readable and
  distinct; an unmodified default request creates or selects the correctly keyed
  256-pixel grid.
• The first request for a model output created before fingerprints were introduced will
  recompute that model in place. The old header is deliberately not treated as proof that
  pre-v0.2.0 revision pins were enforced.
• Plain local output IDs remain basename-based for v0.1 compatibility. Source comparison
  prevents accidental reuse, and manifest preflight rejects duplicate derived IDs.
• Credentials already persisted by v0.1.0 are not rewritten by this release and should be
  rotated separately if the store was shared.

Tracked by #6; this release resolves #2, #3, #4, #5.