Elasticsearch metrics

- removed Prometheus metrics stuff
- removed .cdn-util (might get added back later???)
- renamed and added a lot of configuration variables
- added Elasticsearch metrics collection in metrics/
- changed structure of metrics to include more data
    - GeoIP country
    - response status code
    - object type accessed
    - wildcard-stripped hostname (like before)
- metrics are collected as records per-request rather than a counter
  per-domain

For some reason I forgot to do this or assumed it was done
automatically.

This might be a breaking change for clients expecting a permanent (301
Moved Permanently) status code on redirect objects.

Not to worry, this was logging information that was already logged by Prometheus. No wildcard prefixes were logged.

Via configuration file only, supports wildcards at the beginning of
domains. Useful for anonymizing collected metrics for wildcard domains.

- Add /.cdn-util endpoint (protected by key)
- Add Prometheus metrics (disabled by default) under /.cdn-util
- Replace configuration loader with Viper, config files are now
  supported
- Rename some config flags + environment variables

- object.type == 1 now means "redirect"
- object.long_url is not object.dest_url
    - `ALTER TABLE objects RENAME COLUMN long_url TO dest_url;`
- ensure that 301 with `Location` header is returned on non-preview
  redirects
- set major version to 0, as breaking changes such as this can be
  introduced at any time