Add record-metrics command to Horizon

[bartekn]

Many Horizon admins don't connect Prometheus to scrape Horizon metrics meanwhile checking metrics is essential to debug all kinds of issues. To help debug issues without asking to start Prometheus instance a new command can be added to Horizon that will record /metrics every 15 seconds for defined time (ex. 10 minutes). Then, results are zipped and can be sent to Horizon developers. Horizon developers can start a local Prometheus instance and scrape recorded metrics, then connect them to Horizon instance.

@stellar/horizon-committers thoughts? The solution is not ideal: scraping recorded metrics can take 10 minutes (we can automate it using docker image) but this will give us full picture of a given Horizon instance operations.