fix(migrations): rebuild legacy label index CONCURRENTLY on rollback (MUL-4479) (#5357)

The 162_resource_labels down migration recreated issue_label_workspace_name_lower_idx
with a plain CREATE UNIQUE INDEX, which takes a blocking lock on the existing
issue_label table during a rollback to v0.3.43 and violates the online-migration
rule that every CREATE INDEX must be CONCURRENTLY.

CREATE INDEX CONCURRENTLY cannot run in a transaction or a multi-command migration,
so the rebuild is split out of 162.down:
- 171.down now rebuilds the legacy index as a single CREATE UNIQUE INDEX CONCURRENTLY
  (the natural inverse of 171.up, which drops it).
- 174 (new, no-op up) deletes the agent/skill label rows in its down so they are gone
  before 171.down rebuilds the workspace-wide unique index. Down migrations apply
  high->low, so 174.down -> 171.down -> 162.down runs in the required order.
- 162.down keeps only the transaction-safe structural teardown.

Validated on an isolated Postgres: full up->down->up chain with a colliding
issue/agent label pair; the concurrent rebuild succeeds after the rows are deleted,
restores the valid pre-162 unique index, and a negative control (rebuild before the
delete) fails on the exact collision.

Co-authored-by: J <j@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>

server/migrations/162_resource_labels.down.sql

@@ -1,16 +1,21 @@
+-- Revert issue_label to its pre-162 issue-only shape and drop the resource
+-- label junction tables. The two rollback steps that each need their own
+-- transaction live in earlier down migrations, because a single migration
+-- file cannot mix DML/DDL with CREATE INDEX CONCURRENTLY:
+--   * 174.down removes the agent/skill rows, and
+--   * 171.down rebuilds the legacy workspace-wide unique name index
+--     CONCURRENTLY.
+-- Both run before this file (down migrations apply high->low), so the
+-- non-issue rows are already gone by the time we drop resource_type here.
 DROP TABLE IF EXISTS skill_to_label;
 DROP TABLE IF EXISTS agent_to_label;
 
+-- Defensive: a short-lived pre-release deploy may have applied the original
+-- 162 that created these indexes inline. On the normal down chain 167/168
+-- already dropped them CONCURRENTLY, so these are no-ops.
 DROP INDEX IF EXISTS issue_label_workspace_type_idx;
 DROP INDEX IF EXISTS issue_label_workspace_type_name_lower_idx;
 
--- The pre-162 schema can only represent issue labels. Resource-scoped rows
--- must be removed before restoring its workspace-wide unique name index.
-DELETE FROM issue_label WHERE resource_type <> 'issue';
-
 ALTER TABLE issue_label
     DROP COLUMN IF EXISTS description,
     DROP COLUMN IF EXISTS resource_type;
-
-CREATE UNIQUE INDEX issue_label_workspace_name_lower_idx
-    ON issue_label (workspace_id, LOWER(name));

server/migrations/171_drop_legacy_label_namespace_index.down.sql

@@ -1,3 +1,10 @@
--- Migration 162 restores the legacy index after it removes non-issue labels.
--- Recreating it here could fail while those rows still exist during rollback.
-SELECT 1;
+-- Inverse of 171.up, which drops the pre-162 workspace-wide unique name
+-- index. Recreated as a single CREATE INDEX CONCURRENTLY statement so the
+-- rollback stays online-safe (it cannot be combined with other statements
+-- or run inside a transaction block).
+--
+-- The agent/skill rows that would violate this workspace-wide uniqueness are
+-- removed earlier in the down chain by 174.down (down migrations apply
+-- high->low), so only issue rows remain when this index is rebuilt.
+CREATE UNIQUE INDEX CONCURRENTLY IF NOT EXISTS issue_label_workspace_name_lower_idx
+    ON issue_label (workspace_id, LOWER(name));

server/migrations/174_legacy_label_index_rollback_prep.down.sql

@@ -0,0 +1,6 @@
+-- Remove the resource-scoped (agent/skill) labels before 171.down rebuilds
+-- the pre-162 workspace-wide unique name index, which cannot coexist with
+-- rows that share (workspace_id, LOWER(name)) across resource types. This
+-- runs first in the down chain (high->low), and resource_type still exists
+-- here because 162.down (which drops it) runs later.
+DELETE FROM issue_label WHERE resource_type <> 'issue';

server/migrations/174_legacy_label_index_rollback_prep.up.sql

@@ -0,0 +1,8 @@
+-- Forward no-op. This migration exists only to sequence a rollback step:
+-- its down removes the agent/skill rows from issue_label before 171.down
+-- rebuilds the workspace-wide unique name index. Because down migrations
+-- apply high->low, keeping this delete in the highest-numbered resource
+-- label migration guarantees the non-issue rows are gone before both
+-- 171.down (index rebuild) and 162.down (drops resource_type). There is
+-- nothing to do on the way up.
+SELECT 1;