nagios: improve predictions

In some cases nagios may be able to predict problems by looking at rates of change (e.g. disk filling up), this is done in a few places by calculating slopes of historic data in RRD nagios graph files, but not always alerted on due to noise. Graphs are now beign done via influxdb and this provides a richer set of analysis functions e.g. a derivative of the moving_average or of the expenential_moving_average may work well, or even one of the algorithms used for monitoring trends in stock markets. See https://docs.influxdata.com/influxdb/v1.8/query_language/functions/

## Acceptance criteria

nagios is able to calculate useful rates of change

No files, tests, or entry points are named. Start by locating the existing RRD slope calculations and the InfluxDB graph-query integration, then compare derivative, moving-average, and exponential-moving-average analyses; done means Nagios calculates useful rates of change and alerts on them without excessive noise.

索引モデルが issue の本文から書いたものです。