Notebook: no way to record a channel re-confirmation after re-ordering a unit's channels #258

Found while fixing PR #257 (merged).

Situation. The images of a published experiment are re-exported with their channels in a different order. The user re-orders that unit's channels in the notebook to match, e.g. ACTIN, DNA, PHASE. Spec §10 correctly reopens that repeat's "same way?" confirmation.

• Spec §4.6 (c) refuses moving a confirmed channel ID to another position (R_CONFIRMED_ID_CHANGED) unless a channel re-confirmation is recorded.
• The notebook records re-confirmations only for removals and renames.

So Save is disabled, with the reason shown. That fails closed, but the user has no way forward in the notebook.

Proposal. A small step on Check and save: when a unit's channel positions changed against the published file, list each moved channel ID and ask the user to confirm the new positions explicitly. Record the re-confirmation (§4.6) only on that click, so publication accepts it. Then:

• the reopened repeat gets today's date;
• every other repeat keeps its date byte for byte.

Also: CLI describe still says a header-only change "must be confirmed again", and --publish refuses it. Since #257 the notebook keeps the date for a header-only change resolved through header_names (§10: only a change of a unit's ordered channel list reopens). Align the CLI wording and behaviour with the notebook.

Tests. A header swap resolved by re-ordering, followed by the explicit re-confirmation, saves with today's date for that repeat only. Without the click, Save stays disabled.