Populate derived expression and frozen measures during deployment

[shangyian]

Summary

Metrics created or updated through /deployments now have their derived_expression and FrozenMeasure rows populated atomically with the rest of the deployment, closing a gap where only the single-node create/update paths scheduled this metadata derivation as a background task.

Bulk-deployed metrics previously landed with derived_expression = NULL, which forced every GraphQL read (see this recent change #2049) to re-run metric component extraction.

The orchestrator now includes a "derive measures" phase inside the deployment savepoint (committed on wet-run, rolled back on dry-run), so derivation shares the atomicity guarantees of the rest of the deployment. The new derive_frozen_measures_bulk batches DB work via MetricComponentExtractor's existing cache params, collapsing what would have been N × (2 + avg measures) queries per deployment into ~2 queries (one to load all target revisions + parent chain, one to pre-fetch existing frozen measure rows by name).

Test Plan

• PR has an associated issue: #
• make check passes
• make test shows 100% unit test coverage

Deployment Plan

[netlify]

✅ Deploy Preview for thriving-cassata-78ae72 canceled.

Name	Link
🔨 Latest commit	0a58501
🔍 Latest deploy log	https://app.netlify.com/projects/thriving-cassata-78ae72/deploys/69eb455093b41d00098c7839