promdump: Implement a "minimal" collection mode

[ionthegeek]

Currently, universes with very large numbers of nodes or very large numbers of tables produce too many metrics for promdump to collect. Attempts to collect the tserver metrics cause Prometheus to hang, crash or return data so slowly that the dump fails due to timeout.

A "minimal" collection mode (similar to the YBA yb.metrics.collection_level => MINIMAL setting) should be implemented in promdump as this should make it possible to collect metrics from these very large Universes.