Flaky Service / ingestion Playwright tests block CI #30588

Problem

Several Playwright E2E tests in the Service / ingestion area are flaky and intermittently fail CI, blocking PRs unrelated to their own logic:

• ServiceAgentsPauseResume.spec.ts — the agents tab depends on isAirflowAvailable, which is set from GET /api/v1/services/ingestionPipelines/status. This spec runs in the chromium lane, which structurally has no Airflow container, so the status endpoint returns non-200 → the agent list renders "Ingestion Scheduler is unable to respond" instead of the agent card → getAgentCard(...) times out. The spec mocks the toggle/list/refetch/progress endpoints but not /status.
• ServiceIngestion.spec.ts (nightly) — expensive real run-history execution and block-level timeout handling made the suite slow and unreliable.
• MLflow / S3 / Snowflake / Superset ingestion filter helpers — filter interactions used stale selectors after the filter controls were reworked.

Goal

Make the Service / ingestion Playwright coverage deterministic and independent of a live Airflow / ingestion container in the fast lanes:

• Mock the Airflow boundary comprehensively (including ingestionPipelines/status) so the agent list renders without a real scheduler.
• Replace expensive real ingestion run-history execution with mocked statuses / API-based fixtures in nightly.
• Update ingestion filter helpers (MLflow, S3, Snowflake, Superset) to target the current filter-pattern controls.

Scope

Test-only changes under openmetadata-ui/src/main/resources/ui/playwright/. No product code affected.