Fix teams_name_unique constraint for Hasura on_conflict

[acmeguy]

Summary

• Migration 1741520400001: Replaces the expression index lower(trim(name)) on teams with a plain UNIQUE constraint. The expression index (from migration 1709836800001) cannot be used as a Hasura on_conflict target, which broke team auto-provisioning in dataSourceHelpers.js. Application code already normalizes names to lower(trim()) before insert, so a plain constraint is sufficient.
• docker-compose.dev.yml: Increases hasura_cli migrations server timeout from 30s → 120s to prevent restart loops when applying large migration sets on cold start.

Context

This was discovered as a cluster incident requiring manual DDL fixes. The expression index silently broke the insert_teams_one(..., on_conflict: { constraint: teams_name_unique }) mutation — Hasura requires an actual constraint, not an expression index, for on_conflict targets.

Cluster fixes this prevents from recurring

Fix	Addressed
teams DDL (expression index → constraint)	✅ Migration
hasura_cli restart loop	✅ Timeout increase
Pre-aggregation external: true patches	❌ User-defined schemas in DB, not code
Version checksum resets	❌ Operational, by-design cache behavior
Hasura metadata restore after untrack	❌ One-time incident, metadata in repo is correct

Test plan

• Applied migration locally — verified pg_constraint shows teams_name_unique as type u (unique constraint)
• Tested Hasura insert_teams_one with on_conflict: { constraint: teams_name_unique } — returns null (no-op) for existing team, no error
• Verified hasura_cli starts and reaches healthy state with the timeout increase

🤖 Generated with Claude Code