feat(ground-study): add execution job prerequisites (#4369)

[d-sorgcodexagent]

Summary

• define strict versioned Python/TypeScript regional-ground execution-job serialization with cross-runtime golden evidence
• add typed monotonic progress, cooperative trial-boundary cancellation, terminal failures, and complete-only scalar-ensemble publication
• add strict bounded React import for Python regional scalar-ensemble results with fatal UTF-8 and adversarial validation
• preserve typed null outcomes and deterministic seed/trial/provenance identities
• update SPEC 1.14.63, campaign manifest, and all canonical handoffs

Qualification boundary

This is prerequisite contract/control/parser work, not a runnable-authority claim. Run remains fail-closed. The audit found that the serialization golden uses synthetic flight digests, physical skid/roll settings and executor revision are not yet bound, teed-shot regional surfaces require explicit launch-origin translation, declared orchestration options need full semantics or rejection, and completed results must be bound back to job/input digests. Those blockers are recorded in the handoffs and SPEC.

Validation

• 435 broader Rate/shared variation, ground, and flight Python tests
• 48 composed focused Python tests
• 20 composed focused React tests
• 8 campaign manifest tests
• TypeScript, ESLint, Ruff, focused MyPy, Vite build, module budgets, diff checks, and documentation governance
• successful pre-existing result output remains byte-identical

Closes no issue; advances #4369, #4273, and #4267.

[dieterolson]

Consolidated into #4409. Closing to reduce CI/CD runner load.