Support writing to multiple TSDB in parallel

We would like the ability to write to multiple TSDB concurrently so that we can switch over from Scylla to Cortex without our end users being impacted.

Generally speaking, we would like to be able to mark one tsdb as write-only, and then once there is sufficient historical data, set the second tsdb to read/write and remove the legacy tsdb.

Bonus points if there's a way to mark a tsdb as "backup" in case of a production failure, and OpenNMS can utilize the backup if the primary fails a health check.