partitioned index does not perform well

[Dieterbe]

the partitioned index, introduced in #1232 seems to have some issues.
we deployed it for a customer and noticed:

1. shards were going offline because instances at times couldn't keep up with kafka, increase their priority, and become non-ready
2. overall CPU usage was much higher
3. CPU usage dedicated to GC in particular, was higher as well

Below screenshots show items 2 and 3, on 2 nodes with the partitioned index enabled, and 1 with it disabled. The nodes at this time were are all ready, contained a similar amount of data (tank.total_points, tank.metrics_active, idx.metrics_active), same ingest rate, and were receiving a similar amount of requests

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.