[cenamiller]

Draft: adds support for more than one ECT case, built up one case at a time.

What this does

ECT settings were hard-wired to a single case (120km, mesoscale_reference). This PR moves the per-case settings into one table, .github/ect-cases.yml, and adds an ect-case input to ect-test.yml and ect-ensemble-gen.yml. An empty input uses the table's default, so existing runs behave as before. An unknown name fails with the list of valid cases.

• .github/ect-cases.yml: the table (archive, summary/restart asset names, durations, perturbation, excluded variables)
• .github/actions/resolve-ect-case + .github/scripts/resolve-ect-case.py: resolve a case once in a workflow's config job and pass the values on as explicit inputs
• run-perturb-mpas and validate-ect take their per-case values as inputs; results gain an ect-case column
• The ECT_* per-case keys leave ci-config.env (PyCECT tag/commit stay)
• The restart cache key now ends in the case name, so the first run after merging rebuilds the spin-up restart once

What this does not do

No new automatic runs. Existing push and pull_request ECT runs are unchanged and use the default case. Other cases are run by dispatch (ect-case input). Automatic targeting and a dropdown of cases are intentionally left for follow-ups.

Cases (added to this PR as separate commits)

Each case needs its reference ensemble (summary + spin-up restart in the ECT release, built with ECT Ensemble Generation), and a passing ECT Test run.

Case	Status
mesoscale_reference (default, unchanged behaviour)	in this PR
sf_noahmp	planned
convection_permitting (+ convection-off variant)	planned
limited-area (CONUS) cases	planned

Test plan

• ECT Ensemble Generation dispatched on this branch (default case)
• ECT Test dispatched on this branch (default case)
• Existing per-compiler ECT runs on this PR pass

🤖 Generated with Claude Code