Implement optimized catchup when wallet backend falls behind tip

[aditya1702]

What

• Add important env flags to tweak the backfill worker size and the batches.
• Implement db flushing so that we can insert ledgers frequently while backfilling - this helps prevent increased RAM usage as we don't hold the ledger data in memory.
• Add a BackfillMode iota variable and do an optimized catchup when wallet backend has fallen behind network tip beyond a threshold.
• Add unit tests for the new functionality.
• Add a CatchupTestSuite that tests stopping the wallet backend ingest container, letting the network go ahead and then doing optimized catchup using parallel workers.

Why

This optimized catchup is helpful because for normal live ingestion, we process ledgers sequentially as they are produced. However in cases where wallet backend is behind the tip, we can intelligently fetch multiple ledgers from the ledger backend at once and process them parallelly (sort of like an intelligent backfill) and speed up catchup.

Known limitations

N/A

Issue that this PR addresses

#429

[aristidesstaffieri]

Would it make sense for there to be periodic check points during live ingestion that can kick off optimized catch ups?
Afaict catchups are only detected at startup.

[aristidesstaffieri]

it could be nice to emit a metric for buffer size here so we can see how it relates to theBackfillDBInsertBatchSize

[aditya1702]

I will add it in the metrics PR: #435