Follow-up from PR #179651 (SQL-3328).

Problem

The aggregateDailyUsage activity's query, SqlInterfaceStatusDao.findUnaggregatedSettledDocsUpTo, filters the SqlInterfaceStatus collection on:

{ dailyStatus: { $exists: false },
  date: { $lte: <yesterday> },
  events.propagation_status: { $ne: "PENDING" } }

None of the three existing indexes can serve this query:

• (clusterId, date) (unique) — leading field is clusterId, so it cannot range-scan on date alone
• (meterUsageId) (unique) — unrelated
• (submittedAt) (TTL) — unrelated

As a result MongoDB performs a full collection scan (COLLSCAN) on every daily activity run. At scale (many clusters, one document per cluster per day) this scan grows unbounded within the 180-day TTL window.

Proposed fix

Add a compound index on (dailyStatus, date) via a new BaseCreateIndexMigration (mirroring CreateSqlInterfaceStatusIndexesMigration), then bring the gov backing databases to parity as that migration already does.

Rationale:

• dailyStatus as the leading field lets the planner seek the "absent" bucket (the query targets docs without a dailyStatus).
• date as the second field supports the $lte range bound.
• events.propagation_status ($ne on a multikey array) is not selective and cannot contribute a tight index bound, so it is left as a residual post-filter.

Acceptance criteria

• New index (dailyStatus, date) created via migration, with explicit name matching live commercial naming conventions (idempotent by name).
• explain() on findUnaggregatedSettledDocsUpTo shows an IXSCAN (not COLLSCAN).
• Idempotency integration test added, matching the existing CreateSqlInterfaceStatusIndexesMigrationIdempotencyIntTests pattern.

References

• PR #179651 review comment (nbagnard + Augment)
• systems/sql-interface/migration/src/main/java/com/xgen/sqlinterface/migration/CreateSqlInterfaceStatusIndexesMigration.java