[Storage] delete_journal in journal_table_v2 uses wrong scan prefix for completion-id index cleanup

[tillrohrmann]

Bug Description

In crates/partition-store/src/journal_table_v2/mod.rs, the delete_journal function has a copy-paste bug on line 309: the second deletion loop (intended to clean up completion-id-to-command-index entries with prefix b"jc") incorrectly scans using the notification-id prefix (b"jn") instead.

Since the first loop already deletes all b"jn" entries, the second scan returns an empty iterator, and completion-id index entries (b"jc" prefix) are never cleaned up. This causes orphaned index entries to accumulate in RocksDB over time.

Impact

• Gradual storage bloat from orphaned b"jc" index entries
• Potential stale lookups if completion IDs are reused (unlikely in practice)

Fix

On line 309, change:

notification_id_to_notification_index.clone(),

to:

completion_id_to_command_index.clone(),

Also, the variable name notification_id_index on line 306 should be renamed to completion_id_index (copy-paste artifact from line 282).

Found During

Code review of #4354 (invoker memory budget). This bug is pre-existing and unrelated to the memory budget changes.