chore(kuttl): add test for failing spark job

[sweb]

Description

Adds two kuttl integration test suites: failure-propagation, which verifies that a Spark application failing at runtime is reported with phase Failed and its driver Pod cleaned up, and graceful-shutdown, which verifies that SIGTERM reaches the executor and driver JVMs so that shutdown hooks run and the driver asks its executors to stop.

Tests for stackabletech/docker-images#1595

Definition of Done Checklist

• Not all of these items are applicable to all PRs, the author should update this template to only leave the boxes in that are relevant
• Please make sure all these things are done and tick the boxes

Author

• Changes are OpenShift compatible
• CRD changes approved
• CRD documentation for all fields, following the style guide.
• Helm chart can be installed and deployed operator works
• Integration tests passed (for non trivial changes)
• Changes need to be "offline" compatible
• Links to generated (nightly) docs added
• Release note snippet added

Reviewer

• Code contains useful comments
• Code contains useful logging statements
• (Integration-)Test cases added
• Documentation added or updated. Follows the style guide.
• Changelog updated
• Cargo.toml only contains references to git tags (not specific commits or branches)

Acceptance

• Feature Tracker has been updated
• Proper release label has been added
• Links to generated (nightly) docs added
• Release note snippet added
• Add type/deprecation label & add to the deprecation schedule
• Add type/experimental label & add to the experimental features tracker

[razvan]

A nit from Claude which I like is that the test should be more appropriately called signal-propagation, but I'll leave it up to you to change it or not.

[razvan]

Comment from Claude which I believe is true. The weekly Jenkins tests set VECTOR_AGGREGATOR='vector-aggregator.default.svc.cluster.local:6000' which enables Vector log collection so budget_seconds might not be enough.

budget_seconds=20 is not safe when Vector is enabled (graceful-shutdown/11-check-executor-shutdown.yaml). The test waits for containerStatuses[spark].state.terminated — container exit, not JVM exit. When enableVectorAgent is true the operator injects _STACKABLE_POST_HOOK = sleep 10; (build/pod.rs:241-251). So the real budget is ~10s + JVM shutdown against a 20s limit and a 30s grace period — very little headroom, and the inline claim "A signalled container is gone in 1-2s" is wrong in exactly the configuration nightly runs in. You can't raise the budget much, since it must stay under the grace period.
Better: key the wait off the log, which you're already streaming — wait for Shutdown hook called (or Driver commanded a shutdown) to appear. That is the direct signal, subsumes the separate shutdown-hook grep, and is immune to the post hook entirely. Worth confirming whether #1595's new trap handling runs the post hook on SIGTERM at all — the whole timing budget hinges on it.