[parashardhapola]

Fixes silent wrong results and lost work in the storage layer.

Changes

• cells.insert(name, series) aligns a labelled Series by cell ID instead of by position, which put annotations on the wrong cells.
• A normalizer defined outside CyteArc is identified by a digest of its code and defaults, so an edited normalizer is not reused; revision 2 of the operations that record one.
• A missing count, countsT, or artifact chunk raises instead of reading as zeros.
• Stores with consolidated metadata no longer break writes.
• Ctrl-C in Jupyter stops background readers and writers before control returns.
• ds.artifacts.prune_incomplete(dry_run=True) finds and removes interrupted artifacts.
• A fresh import is prepared by one process under an atomic claim, and min_features_per_cell applies only on the first writable open, which records it.
• The Harmony 1.2.0 change is registered as a scoped revision; CyteArc no longer replaces other libraries' log handlers; listing, lineage, and argument checks are hardened.

Behaviour changes

• A Series with a non-range index is aligned by label or raises.
• Reopening a store never filters I again.
• A store written by another tool that skips all-zero chunks no longer reads; the published CyteBase stores read normally.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!