[SPARK-51019][SQL][TESTS] Fix flaky test: SPARK-47148 AQE should avoid to submit shuffle job on cancellation

[ashrithb]

What changes were proposed in this pull request?

This PR fixes a flaky test SPARK-47148: AQE should avoid to submit shuffle job on cancellation in AdaptiveQueryExecSuite.

The fix replaces the shallow exception chain search with a recursive traversal that finds the "coalesce test error" message at any depth in the exception hierarchy.

Why are the changes needed?

The test was flaky because the original assertion only checked:

• The top-level exception
• Its direct cause
• Its immediate suppressed exceptions

// OLD (flaky):
val errMsgList = (error :: error.getCause :: error.getSuppressed.toList)
  .filter(e => e != null && e.getMessage != null)
  .map(_.getMessage)

assert(errMsgList.exists(_.contains("coalesce test error")))

But, from race conditions in AQE's concurrent execution, the "coalesce test error" can be wrapped at varying depths depending on:

• Thread scheduling/timing
• Which AQE stage reports the error first
• How exceptions are wrapped during job cancellation

When the error was nested more than one level deep, the assertion failed.

Does this PR introduce any user-facing change?

No. This is a test-only change.

How was this patch tested?

• Ran the test 20+ consecutive times locally with 100% pass rate
• Before the fix, the test would intermittently fail in CI (see linked JIRA)

Was this patch authored or co-authored using generative AI tooling?

No.