test(integration): fix flaky JiT test racing on shared tmp dir cleanup

[mbiernacik]

The two tests in tests/integration/jit.spec.ts run in parallel and shared a TmpDirFixture cleaned up in afterEach, so the test that finished first deleted the other test's project directory. When that happened before the other test's JiT step, the JiT worker couldn't find @dataform/core and the run failed with expected 'FAILED' to deeply equal 'SUCCESSFUL'. BigQuery job history shows this in 16 of 388 CI runs over the last 90 days.

Changes:

• Each test owns its project directory and removes it in finally, so tests can't delete each other's files.
• Datasets get a per-run random suffix and are dropped when the test ends, so concurrent runs (e.g. two CI builds) can't interfere and nothing is left behind.
• The status assertions include the action results, so a failure shows the underlying error.

[apilaskowski]

nit: process.env.TEST_TMPDIR is only set under Bazel. Outside it, path.join(undefined, ...) throws. Maybe process.env.TEST_TMPDIR || os.tmpdir()?

[mbiernacik]

Good point, done.

[mbiernacik]

Heads-up: we're going to close this PR in favour of #2385.

Both PRs fix the same root cause: integration tests reusing fixed BigQuery dataset names, so concurrent runs and tests drop or overwrite each other's data. This PR does it only for jit.spec.ts. #2385 does it across all the integration suites (jit, run, evaluate, bigquery, property_graph, plus the CLI e2e/JiT tests), with shared helpers in testing/bigquery.ts and a cleanup of stale datasets left behind by killed runs. Rather than having two different conventions (makeSuffix here vs makeUniqueSuffix there, which are also incompatible with the stale-cleanup regex), we're standardising on #2385.

The part that's unique to this PR, each JiT test owning and cleaning up its own temp directory instead of using TmpDirFixture(afterEach) in a parallel suite, plus the expectSuccessfulRun diagnostics, is being carried over to #2385 (see the review there). Nothing from this PR is lost. Further work continues on #2385.