Add prometheus metrics

[jdreichmann]

Continuation of #71

The following metrics are available for scraping with prometheus:

• bridge_rooms_total: labeled by protocol and type (dm, group)

• bridge_puppets_total: labeled by protocol, how many puppets exist (aka remote accounts matrix users have linked)

• bridge_remote_users_total: labeled by protocol, how many remote users (sometimes called "ghosts") exist

• bridge_messages_total: labeled by protocol and type, so count(bridge_messages_total) by (protocol, type) is possible

• bridge_matrix_events_total: labeled by protocol and type, the total amount of matrix events processed since startup

• bridge_matrix_event_errors_total: labeled by protocol, how many matrix events resulted in an exception during bridging

• bridge_matrix_event_seconds: histogram by protocol and type of the time spent processing a matrix event to be bridged

• bridge_remote_update_seconds: histogram by protocol and type of the time spent bridging remote events into matrix

Labeling by the protocol id enables you to have a single scrape config in prometheus with many mx-puppet-* bridges as targets, and easily visualizing global bridge activity with f.ex. grafana.

For bride_connected to work, the implementing bridge needs to call trackConnectionStatus on connect/disconnect.