Updates for compatibility with Huber-group-EMBL/romeo

[Artur-man]

• export ImageArray class
• remove axes slot and parse axes ordering from scale
• update scaling strategy from 1/s or s

[Artur-man]

@claude review

[Artur-man]

@claude review again

[Artur-man]

@claude review, check again but make it short

[claude]

Claude finished @Artur-man's task in 2m 43s —— View job

---

Review

• Gather context (diff, changed files)
• Review code changes
• Post review

Went through the diff (origin/devel...HEAD) in detail. Overall this is a clean, well-tested change — the three goals (export ImageArray class, derive axes from scales instead of a separate slot, flip the scale convention to 1/s) are implemented consistently across R/ImageArray.R, R/Validity.R, R/transformations.R, and R/manipulation.R, and the new tests in test-axes.R/test-scales.R cover the permutation and validation logic well.

A couple of minor points, nothing blocking:

1. R/AllClasses.R:34 — ImageArray class is now @exportClass'd but still tagged @keywords internal, so it won't appear in the reference index. Since the whole point of exporting is downstream use by romeo, worth double-checking that's intentional rather than an oversight.
2. R/ImageArray.R axes<- — the replacement only permutes the scales metadata (via scales(object)[value]), it does not reorder the underlying array data in @levels. That matches the old behavior (metadata-only), but could be worth a doc note since it's a common point of confusion (axes labels vs. physical dimension order).
3. Style nit: R/Validity.R dropped braces from several if (...) stop(...) statements — fine as-is, just flagging since it's a slight style shift from the rest of the file.

No correctness issues found in the scale-direction flip (createListFromEBImage, .magick_resize_scale, .get_scales_from_dim, .get_scales_from_nlevels, affine/scale/crop in transformations.R/manipulation.R) — all consistently updated to the new 1/s convention and backed by matching test updates.

Note: no R environment available in this sandbox, so this is a static review only (couldn't run devtools::test() or R CMD check).