refactor: align persistence layer with UUID identifiers and Drizzle schema typing

[heitorrsdev]

Description

This PR consolidates three structural improvements: stabilization of the test database container, standardization of all primary and foreign keys as UUID, and removal of manual row typings in favor of Drizzle-inferred types.

The changes improve architectural consistency, reduce schema duplication, and strengthen alignment between database and application layers.

---

What was done

• Stabilized test database environment:

  • Added healthcheck to the Postgres test container
  • Disabled restart policy to prevent flaky behavior
  • Removed temporary debug log from concurrency test

• Standardized identifier strategy:

  • Migrated primary and foreign key columns from TEXT to UUID
  • Recreated affected constraints and exclusion rules
  • Added ADR documenting the UUID decision and its implications

• Reduced type duplication and coupling:

  • Replaced manual row interfaces with .$inferSelect / .$inferInsert
  • Removed explicit row typing in reservation repository
  • Aligned mappers and tests with schema-derived types

---

Why

• The test container configuration could lead to unstable or misleading test runs
• Identifiers were semantically UUID but structurally defined as TEXT, creating inconsistency
• Manual row types duplicated schema definitions and increased maintenance cost
• Schema-derived typing strengthens architectural boundaries and reduces drift between layers

---

Notes

• No business behavior was modified
• Database schema change is safe (no persisted data exists)
• All changes are covered by existing integration tests
• The UUID decision is formally documented under docs/adr