Separate reaping from ingestion

[2opremio]

As part of #196 , it could be that we need to reap a large region of history on startup (due to a change in the history window parameter).

Thus, reaping can take a long time. It shouldn't be part of ingestion since it could block it.

As part of this task we should create a reaper which reaps a range of ledgers. The orchestration of what ledgers to reap should be part of #202

The reaper should (together with the orchestrator) to make sure no inconsistencies are left on the database. In particular it should do it forward form the last ledger so that no gaps are left if the reaping process stops half-way.