Record completeness and segmentation snapshot provenance in mask_relationships.json #166

Problem

A mask scan can legitimately run while segmentation is incomplete because missing masks are warnings, not errors. However, mask_relationships.json currently looks like a final experiment-level description even when only part of the experiment has masks.

There is also a wording/provenance mismatch: the notebook says Describe the masks describes the recorded run, but the scan performs a fresh preflight and reads the masks currently present. A valid mask may therefore have changed or been added since the preprocessing manifest.

For live/incremental segmentation we need to know exactly what snapshot the report describes.

Suggested simple implementation

Make the report self-describing.

Record at least:

• total analysed experiment FOVs;
• FOVs with any masks;
• per mask type: masks present / masks expected / missing;
• for each pairwise relationship, how many FOVs had both mask types and contributed evidence;
• whether the segmentation snapshot is partial or complete;
• enough mask fingerprint/digest provenance to identify the exact mask snapshot used.

Use one clear policy in the notebook/CLI:

• either scan the current segmentation snapshot and say so explicitly; or
• require the exact masks recorded by the preprocessing run.

For the intended progressive-segmentation workflow, scanning the current snapshot with explicit provenance is probably the simpler and more useful model.

Acceptance

• A report from 60/100 segmented FOVs cannot be mistaken for a complete 100-FOV report.
• Missing masks are visible as completeness information, not only warnings outside the report.
• Relationship evidence states the number of FOVs that actually contributed.
• A mask added/changed after the preprocessing run is explicitly identifiable in the scan provenance.
• The notebook wording matches the implemented snapshot policy.
• A final scan after segmentation completion is unambiguously marked complete.

Relevant code: mask_scan.MaskReport, run/mask_step.py, notebook mask-step text.