fix(meta): unregister aborted initial job tables

[Li0k]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

• Related to bug: cancel job doesn't notify hummock correctly #25466
• Split from fix(meta): cleanup dirty recovery hummock tables #25592

This PR is scoped to DDL abort cleanup for creating streaming jobs. It does not change per-database dirty-job recovery cleanup; that path is fixed separately in #25592.

Bug:

• When creating a streaming job fails, create_streaming_job_inner aborts the creating job with is_cancelled = false.
• DROP/CANCEL of an Initial job also aborts through the same catalog helper.
• Before this PR, these DDL abort callers could delete catalog rows without running the existing dropped-table cleanup path, so Hummock table ids that had already been registered could remain stale.
• The real SQL/simulation repro uses per-database recovery only to make the foreground create fail. After the create fails, the stale-id cleanup responsibility is in the DDL abort path fixed here.

Fix:

• Return an AbortCreatingStreamingJobResult from try_abort_creating_streaming_job, including the aborted job's state table ids.
• Collect table ids from internal table catalogs, Fragment.state_table_ids, and the MV table catalog. Fragment.state_table_ids is needed because the SQL/simulation repro registers table ids through fragment metadata that are not fully represented by get_internal_tables_by_id.
• Route DDL create failure and DROP/CANCEL of Initial jobs through cleanup_dropped_streaming_jobs, so Hummock unregister and Hummock/compactor notification happen through the existing dropped-table cleanup path.
• Keep replace-job abort handling and live Background/Creating recovery unchanged.

Not covered by this PR:

• Dirty creating-job cleanup during per-database recovery. That is fix(meta): cleanup dirty recovery hummock tables #25592.

Testing

• cargo fmt --all -- --check
• git diff --check
• cargo check -p risingwave_meta --lib
• cargo test -p risingwave_meta test_abort_initial_materialized_views_unregister_hummock --lib -- --nocapture
• ./risedev sit-test test_per_db_recovery_abort_failed_foreground_ddl_unregisters_hummock_tables --no-capture

Checklist

• I have written necessary rustdoc comments.
• I have added necessary unit tests and integration tests.
• I have added test labels as necessary.
• I have added fuzzing tests or opened an issue to track them.
• My PR contains breaking changes.
• My PR changes performance-critical code, so I will run (micro) benchmarks and present the results.
• I have checked the Release Timeline and Currently Supported Versions to determine which release branches I need to cherry-pick this PR into.

Documentation

• My PR needs documentation updates.

Release note

Fixed stale Hummock table ids that can remain after aborting a creating streaming job in the DDL path.