convert: accept a stated page reshape (e.g. --pages CZ=2x3) and/or reorder copies into the common order (plan Q73) #144

Follow-up of plan §12 question 73 (F8 of the cross-brick test of 3 November 2026). Post-matrix: not for v0.4.

The owner decided on 3 November 2026 that v0.4 has neither of these; this issue keeps them for after the matrix run.

Today (v0.4.0rc5)

• F8a: convert keeps each image's stored order of dimensions (contract §17). inconsistent_dimensions and inconsistent_mapping say that named-axis copies do not solve them.
• F8c: a plain stack whose pages interleave several dimensions (IYX, no shape metadata) cannot be converted. A reshape is not a role, and the axis spec cannot state one. The user guide says to save the image again with its dimensions (for example as a Fiji hyperstack).

Options to decide

1. A stated page reshape, for example --pages CZ=2x3: the user states how the pages split into dimensions and in which order (C slowest, Z fastest). The reshape is recorded in the conversion record and checked against the page count. It is never guessed from the page count.
2. Reorder copies into the experiment's common order: a pixel-exact transposition under the named roles, so that the copies share one order. This changes the copy axes in §17 and the conversion record (version bump).

Either option needs:

• contract §17 and the conversion record updated;
• the CLI and the notebook's Make named-axis copies card;
• tests showing pixel-exact copies, a refusal of a reshape that does not match the page count, and that nothing is guessed.

Refs: docs/plan.md §12 Q73; PR #139.