fix: resolve database query antipatterns in GraphQL resolvers

[nimish-ks]

Summary

• Cache repeated OrganisationMember lookups in OrganisationType — 5 identical queries per request reduced to 1
• Eliminate N+1 .exists() loops in AppType.resolve_environments, resolve_app_environments, ServiceAccountType.resolve_app_memberships — replaced with upfront values_list set lookups
• Batch environment queries in OrganisationMemberType.resolve_app_memberships — one query for all apps instead of one per app
• Add select_related for organisation and user foreign key access to avoid lazy-load hits
• Wrap multi-write mutations in transaction.atomic() — CreateEnvironment, UpdateMemberEnvScope, CreateServiceAccount
• Fix race condition in SwapEnvironmentOrder with select_for_update() + transaction.atomic()
• Bulk delete service account handlers instead of loop-based individual deletes
• Collapse redundant queries in PhaseLicenseType — .exists() + .get() + extra Role lookup → single .first() with join
• Set-based access filtering in syncing queries instead of per-item user_can_access_app calls

Test plan

• Verify environment listing works correctly for users with partial environment access
• Verify service account app memberships resolve correctly
• Test environment creation, member scope updates, and service account creation still work atomically
• Test swapping environment order under concurrent requests
• Verify organisation-level sync listing returns correct results
• Verify license activation and owner resolution on EE