Split AREG into three tools, fix what the splits left stranded, and repair CI

[Jules-GP]

What this does

Completes the AREG split (AREG_CBCT / AREG_IOS / AREG_IOSCBCT over a shared
common), and fixes what both this split and the earlier ALI one left behind.

The defects the splits left, and why nothing reported them

Splitting a tool leaves its old tests/test_run.py importing the pre-split
package. pytest reports that as one collection error, so the whole file stops
running: 84 tests for AREG, 16 for ALI. One error reads as one problem.

Behind it:

• AREG_IOS asked the supervisor for ALI, renamed ALI_IOS by the split.
  The mucogingival mode could never have run. The server's startup check missed
  it because describe.py scanned sup.run("X") and not
  tools.require(sup, "X", ...) — the hole that function's own docstring warns
  against. Both are fixed, with tests.
• ALI_CBCT's error message named cbct_regions, an argument the split had
  renamed to regions, so it told callers to fill a field the schema does not
  have.
• AREG's error messages pointed at GET /tools/AREG/data, a 404 since the
  split.
• ASO reached split_scan_extension through the CBCT pipeline rather than
  its own scans module.

CI

• The matrix was built with find -mindepth 2 -maxdepth 2, which stopped
  matching when ALI and AREG became grouping folders. CI has been testing six
  tools and silently skipping the five the splits produced — which is where the
  stale tests were found, months later, by hand. It now walks depth 2 and 3 and
  keys on [tool.sadt], the same rule the registry and the deployment
  Dockerfile use. That also drops _template, which CI tested and no server
  serves.
• actions/checkout and astral-sh/setup-uv moved to the lowest majors running
  on Node 24, clearing the deprecation warning.

Verification

Every suite, run in its own interpreter:

tool	tests		tool	tests
AMASSS	51		ALI_CBCT	43
ASO	82		ALI_IOS	12
Crown_Seg	22		AREG_CBCT	25
Surg_Mov_Pred	24		AREG_IOS	53
Batch_Dental_Seg	24		AREG_IOSCBCT	1
scripts	64

All ten tools were also run end to end over HTTP against the deployment image,
through the Slicer client, on real data — including the three-level supervised
chain (AREG_IOSCBCT to ASO to ALI_CBCT).

AREG_IOSCBCT keeps only one test of its own; the rest of the merged suite
belonged to the other two. Its own argument rules deserve a suite and do not
have one yet.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.