Improve database index management during ENSIndexer restarts

[tk-o]

Currently, when ENSIndexer instance restarts, it will always go back into the backfill and drop existing database indexes.

https://github.com/ponder-sh/ponder/blob/ponder%400.16.6/packages/core/src/database/index.ts#L950-L964

This is a huge issue for ENSIndexer Alpha instance, where creating database indexes can take multiple hours (i.e. for Railway project). Ponder internal strategy for managing restarts could be improved by avoiding the need to re-create large database indexes.