File-controlled scrapers that can run arbitrary queries

[rcrowley]

This patch adds --extras.file and --extras.refresh-interval to mysqld_exporter in order to allow easy addition of arbitrary metric-gathering queries without restarting mysqld or worse.

Usage:

cat >extras.yaml <<EOF
---
- metric: foobar
  query: SELECT 47 AS foo, 48879 AS bar;
- metric: bazquux
  query: SELECT 3.14159 AS baz, 2.71828 AS quux;
...
EOF
mysqld_exporter \
    --extras.file /path/to/mounted/config/map/extras.yaml \
    --extras.refresh-interval 1m

When scraped, these metrics will be among the output:

# HELP mysql_extra_bazquux Extra metrics from SELECT 3.14159 AS baz, 2.71828 AS quux;
# TYPE mysql_extra_bazquux gauge
mysql_extra_bazquux{column="baz"} 3.14159
mysql_extra_bazquux{column="quux"} 2.71828
# HELP mysql_extra_foobar Extra metrics from SELECT 47 AS foo, 48879 AS bar;
# TYPE mysql_extra_foobar gauge
mysql_extra_foobar{column="bar"} 48879
mysql_extra_foobar{column="foo"} 47

The YAML file is not intended to be stuffed in the filesystem and forgotten about. Instead, it should be served via a Kubernetes config map. To update the metrics and queries that it scrapes, update the config map and wait longer than about the --extras.refresh-interval plus one minute (empirically how long config map updates take to become visible).

ENOENT and EOF on the --extras.file are not considered errors. It might exist next time! This also means that deleting or emptying the file is a valid way to stop collecting these metrics.

Post merge, here's what comes next:

1. Figure out how to build this container image.
2. Test this patch on my pet turtle.
3. Make a release plan and do it.