feat(skills): author the ac_* CTI calibration skill set (Phase 2)

[Jammy2211]

Phase 2 of the autocti_assistant epic (PyAutoBrain#136). Grows the lightweight
seed into a working assistant: nine skills covering the CTI calibration
pipeline end-to-end, each grounded in a validated autocti_workspace script and
verified against the live stack.

The skills

Skill	Covers	Grounded in
ac_setup_environment	install incl. arcticpy (the #1 failure)	PyAutoCTI:AGENTS.md, smoke_install.sh
ac_workspace_navigation	the map: two geometries, shared stage pipeline	autocti_workspace tree
ac_simulate_dataset_1d	simulate a 1D CTI dataset with a known trap model	dataset_1d/simulators/start_here.py
ac_simulate_imaging_ci	simulate a 2D charge-injection image (Clocker2D)	imaging_ci/simulators/start_here.py
ac_mask_and_extract	mask FPR/EPER; layout.extract.fpr/eper	dataset_1d/{modeling,extract}
ac_fit_cti_model	compose + fit, incl. the AnalysisFactor/FactorGraphModel joint fit	dataset_1d/modeling/start_here.py
ac_correct_cti	clocker.remove_cti for correction + diagnosis	dataset_1d/correction/start_here.py
ac_plot	the autocti.plot function API (no MatPlot/Output object)	scripts/plot/start_here.py
ac_load_results	result objects + the aggregator (ac.agg.CTIAgg)	dataset_1d/results/

Wired into .claude/skills/ (symlinks), indexed in a rewritten
skills/README.md — the old index was the autolens catalogue with al_→ac_
substitution, listing 45 skills of which only 13 files existed; it now honestly
describes the CTI catalogue and points at PENDING.md for the growth queue.

Grounding beat the prompt's assumptions twice

• The aggregator is ac.agg.CTIAgg, not Dataset1DAgg as the seed prompt
  guessed (both exist; CTIAgg is what the scripts use).
• The plot scripts live at scripts/plot/, not per-geometry — my first draft
  cited scripts/dataset_1d/plot/ and the citation check caught it.

Validation — every wiki-currency check green

Run with audit_skill_apis.py against the live CTI stack:

--scope all         PASS   (0 broken / 33 symbols)
--check-citations   PASS   (0 missing / 27 citations)
--lint-idioms       PASS
--check-provenance  PASS
--check-version     PASS   (wrote wiki/core/api_audit_baseline.json,
                            pinning the 4-module CTI stack — no autogalaxy)

The audit earned its keep three times: it caught aplt.Output and
analysis + analysis written in prose to say they're gone (rephrased so the
parser doesn't read them as live symbols), a wildcard aplt.subplot_* token, and
the wrong plot-script paths.

Also fixes the pre-existing _style.md lensing-path citation
(scripts/imaging/ → scripts/dataset_1d/modeling/), the last seed content leak
the earlier PR flagged.

Expected CI

wiki-currency should now go green — the reason it was red on the seed-infra
PR (no baseline) is resolved, and every check above passes locally against the
same main source the CI installs. boundary skips (not a clone reference).

Part of #136 (Phase 2). Wiki (Phase 3) and end-to-end demonstrations (Phase 4)
still to come; the newborn stays private until then.