[Jack-Coutts]

Why

After a tR change, plots of one compound could be centred on different tRs, even when every sample had the same tR. Apply re-extracted a sample only when the new integration window fell outside its stored EIC, with 0.001 min of slack. A tR change inside the stored data left the plot where it was. Sub-second differences in each file's first scan decided which samples moved. A user report showed 15 of 25 Sucrose samples re-centred and 10 left on the old tR.

What changed

• samples_needing_recut reads each sample's stored EIC from the database. Apply re-extracts every sample whose EIC is missing, is more than one scan from its tR, or does not cover tR minus the left offset to tR plus the right offset.
• The per-sample bounds cache and the coverage helpers are deleted. That cache also stored bounds under the wrong sample (Integration window stores data bounds under the wrong sample #188).
• Apply finishes without a "Regeneration Complete" box after it re-extracts, as it does when nothing is re-extracted. Update tR Window and the other explicit regenerations still show it.

Scope

• Session import still uses its own coverage check. The follow-up on the session stack switches it to this rule.
• An offset-only change still re-extracts nothing unless the offsets outgrow the stored data.

Tradeoffs

Re-extracting on every tR change costs about 30 ms per 107 MB sample once #265 lands, and about 340 ms without it. Extracting a wider margin and only re-centring on large moves would keep small edits instant, but it keeps the coverage check this PR removes.

Blast Radius

Every tR change made with Apply now re-reads the CDF files for the samples it applies to. Compounds near the start or end of a run re-extract on every Apply, because their traces are clipped. The results are identical.

Verification

• uv run pytest: 513 passed. The new tests cover a trace off-centre but covering the offsets, a trace narrower than the offsets, a missing trace, a trace clipped at the run start, and a trace with a scan gap.
• Drove the real MainWindow offscreen on the 30-sample labelled bench. Plotting Pyruvate and applying tR 6.93 (left offset 0.035, right offset 0.05) to all tiles: on main, all 30 tiles stay centred on 7.088. On this branch, all 30 are centred on 6.929, with no message box. With tR unchanged, nothing is re-extracted.

Fixes #264
Fixes #188

[Jack-Coutts]

Replaced by #302.