Update Prometheus flags

Prometheus needs to be configured for how much memory is used for buffering metric data.

• Memory chunks are 1kb in size.
• The minimum memory needs to be 2x as many chunks as there are unique metrics time-series.

See: https://prometheus.io/docs/operating/storage/#memory-usage

• Update memory allocation controls
  • Use 10k + 2% of memory for memory-chunks.
  • Set half of that for max-chunks-to-persist.
• Set default for chunk encoding version to 2
  • Trades some CPU time for increased compression.