fix(admin): re-key legacy plans before seeding, tolerant of a prior partial boot

Boot ran SeedPlans before Backfill, so the fresh (self_hosted, professional)
seed row was inserted before pass 3 tried to rename the legacy self_hosted TIER
row into it — colliding on deployment_tier_unique and failing boot.

The plan re-key moves to MigrateLegacyPlans, called before SeedPlans so the
rename lands first and the seed no-ops on it. It is also self-healing: on a
database a crashed boot already seeded (self_hosted, professional) into, the
legacy row can no longer be renamed onto it, so it is dropped instead — its
instances are re-tiered and re-entitled from the surviving professional row.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

admin/cmd/main.go

@@ -64,6 +64,13 @@ func main() {
 		idxCancel()
 		log.Fatalf("indexes: %v", err)
 	}
+	// Legacy plans are re-keyed BEFORE the seed, so the seed's fresh
+	// (self_hosted, professional) row cannot collide with the legacy self_hosted
+	// row's rename on deployment_tier_unique.
+	if err := models.MigrateLegacyPlans(idxCtx); err != nil {
+		idxCancel()
+		log.Fatalf("migrate legacy plans: %v", err)
+	}
 	if err := models.SeedPlans(idxCtx); err != nil {
 		idxCancel()
 		log.Fatalf("plan seed: %v", err)