Make collectd WriteQueueLimitHigh WriteQueueLimitLow configurable

[Miles-Garnsey]

Collectd acts as a metrics resevoir and serves metrics to Prometheus via json on an exposed port.

The number of metrics stored in collectd appears to be governed by two settings WriteQueueLimitHigh WriteQueueLimitLow, (set equal by default). When the resevoir is full, incoming metrics are dropped.

These options should be configurable so that (especially in clusters with large numbers of tables) collectd can be configured to store more metrics.