fix(db): drop the unused plain env_builds status index

[ValentaTomas]

Summary

idx_env_builds_status (plain btree on raw status) has a negligible number of lifetime scans in pg_stat_user_indexes. Status-filtered access runs on the purpose-built indexes instead (idx_env_builds_env_status_created, idx_env_builds_team_status_group, idx_env_builds_team_status_pagination, and the covering index including status_group).

env_builds is the most write-amplified table in the registry (0% HOT updates, 8 indexes), so every dropped index removes real WAL and dead-tuple volume from each build insert/supersede.

Draft on purpose: requesting an owner's confirmation that nothing planned depends on a full-table raw-status scan path before this lands.

[cursor]

PR Summary

Medium Risk
Dropping an index on a ~330M-row hot table can still block or slow workloads if anything still relies on full-table status scans; the down path is a long manual concurrent rebuild.

Overview
This removes the plain btree index on raw env_builds.status because it is barely used while every build insert still maintains it; status-filtered queries are expected to use the newer composite and status_group indexes instead. The migration uses DROP INDEX CONCURRENTLY with a one-hour statement_timeout on up and restores the migrator’s three-hour default afterward; down clears any invalid leftover index, rebuilds concurrently with no timeout for manual rollback, then restores the three-hour default.

^{Reviewed by Cursor Bugbot for commit 428983f. Bugbot is set up for automated code reviews on this repo. Configure here.}

[ValentaTomas]

closing for now, will revisit