Add continuous query for ETag caching events

See merge request !8

Add license

See merge request !7

These queries calculate the allocated memory per Rails/Grape/Sidekiq
controller/worker per minute. This data includes the 95th percentile,
the 99th percentile, the mean, and the max value.

InfluxDB decided that as of 1.0.0 it would be a good idea to rename the
default retention policy from "default" to "autogen" and then hardcode
this name all over the place. This means that we not only have to handle
this when creating retention policies, we also have to make sure all
queries use the right policy name for input measurements.

Fixes #1

Having put some more thought in this it doesn't make sense to sum the
total durations as this can be influenced heavily by lots of fast
requests.

This reverts commit 8e2c18f9.

Downsample duration sum (to calculate total consuming time)

See merge request !6

Downsample memory_usage max values (rails / sidekiq)

See merge request !5

Apparently running a query with "RESAMPLE EVERY 30m FOR 1h" _without_
an aggregate function results in InfluxDB throwing an error. This in
turn leads to it retrying the same CQ every second, because if it
doesn't work the first time it will _probably_ work the 12039812093th
time.

Added queries for tracking deployments

This is for https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/5320

See merge request !2

This ensures the deployments are copied into "forever.deployments" which
has no expiration time (unlike "default" which expires after one hour).

Add cache read hit / miss counters continuous queries (Rails/Grape/Sidekiq)

See merge request !3

Downsample Sidekiq Queue Duration Max

See merge request !4

Sidekiq queue duration timings

See merge request !1