fix(sync): run warehouse-backed syncs on postgres

[jonathanhaaswriter]

Summary

• switch direct/API/scheduled sync paths to the generic warehouse interface and make Postgres the default warehouse when Postgres DSNs are configured
• add SQL dialect helpers so sync upserts, relationship persistence, schedule storage, and change-tracking tables work on Postgres-backed warehouses
• update warehouse/status health surfaces and regression tests to cover the Postgres-backed sync runtime

Validation

• go test ./...
• make verify (race tests completed; remaining verify targets run manually because the installed golangci-lint binary targets Go 1.25)
• GOTOOLCHAIN=go1.26.1 go run github.com/golangci/golangci-lint/v2/cmd/golangci-lint@v2.8.0 run --timeout 5m ./...
• make api-contract-compat
• make cloudevents-contract-compat
• make report-contract-compat
• make entity-facet-contract-compat
• make agent-sdk-contract-compat
• make graph-ontology-guardrails

[jonathanhaaswriter]

I think there are two issues here:

• JOB_DATABASE_URL now implicitly flips the warehouse backend to Postgres, so setting it just for app-state/jobs can silently disable Snowflake unless WAREHOUSE_BACKEND is also forced
• the post-sync toxic-combination scan is now routed through application.Warehouse, but the implementation still uses Snowflake-specific SQL, so Postgres-backed syncs will just warn/skip that analysis

Can we decouple warehouse selection from JOB_DATABASE_URL and either port or gate the toxic-combination scan for Postgres?