Serving SRV targets and getting all nodes in single request

[adnull]

As it stated in #773 there is no chance to monitor cluster nodes with the only mongodb+srv url specified.
According to the specification, direct connection couldn't be used with a cluster, specified as a SRV domain

This update adds

• --split-cluster option, which:
  • extracts mongodb+srv domain into cluster connection string containing nodes from DNS server
  • splits all cluster connection string into separate targets that can be queried by "target" parameter of multitarget endpoint
• overall target endpoint "/scrapeall" which outputs all the targets metrics with a "instance" label containing nodename as a host:port pair (or only host if no port specified)

Combining all that features we can specify cluster url as mongodb+srv domain and monitor all the nodes metrics and states in one request by querying /scrapeall endpoint