Add a setting to activate read-repairs

[iksaif]

Also set compaction options accordingly when doing it. Read repairs will write in the past which means that we need to set these correctly:

• tombstone_threshold (something high, like 80%)
• tombstone_compaction_interval (something higher than TTL + gc_grace_period)

[iksaif]

This should probably only be set once we've fixed issues on Cassandra with overlapping fully expired sstables.

[iksaif]

These is a patch for overlapping sstables available on https://issues.apache.org/jira/browse/CASSANDRA-13418

[iksaif]

Patch is in 3.11.1, so we're good. Now we need to at least write doc on how to configure read repairs.