test: reproduce config-sync dropping remote cluster additions

[ilyakuz-db]

Superseded by #6631, which uses the same commits on a branch in databricks/cli.

Changes

Add an acceptance test showing that #6531 makes bundle config-remote-sync --save silently ignore a remotely added Spark setting on a job cluster with a policy.

Why

The test's policy fixes only custom_tags.PolicyTag; it leaves spark_conf unrestricted. After deployment, the test adds spark.sql.shuffle.partitions: "7" to the job's task cluster through the Jobs API. Config-sync should copy this change into databricks.yml:

+            spark_conf:
+              spark.sql.shuffle.partitions: "7"

Main produces this diff. With #6531, config-sync reports No changes detected and leaves YAML unchanged. The new rule skips remote-only additions whenever the cluster has policy_id, without checking whether the policy defines the field.

This test-only PR is stacked on #6531 and intentionally fails until that regression is fixed.

Tests

• Main (9caaff0c): the exact acceptance test passes locally with deployment history off/on and on e2-dogfood with history off. Snapshot generated by running main.
• Do not report cluster fields supplied by a cluster policy as drift #6531 (8aeb76c4): the same local and cloud runs fail because the YAML diff is missing.
• Main's cloud runs with deployment history enabled were blocked by backend HTTP 500s during deploy/cleanup.
• Formatting, checks, and lint pass. The full suite stopped on unrelated libs/patchwheel failures (ModuleNotFoundError: myproj).

This PR was written by Codex.