soroban-rpc: Add opt-in remote metrics push so we can collect usage data

[paulbellamy]

What problem does your feature solve?

As SDF, we would like to know how the deployed soroban-rpc instances are running. What kind of resources they have available, and what sort of load they are dealing with.

What would you like to see?

An opt-in option for operators, which would periodically push their metrics to a remote target (e.g. a prometheus pushgateway server we're running to collect them).

We probably want to use: https://pkg.go.dev/github.com/prometheus/client_golang/prometheus/push

We should be able to tell:

• Total number of RPC instances by
  • network (futurenet, testnet, mainnet)
  • locale
• Top endpoints by request signature (ie: /getEvents, /getLedgerData)
• Average request latency by endpoint
• Total unique Users per RPC instance
• Top error codes and failures
• Pseudorandom identifier that would help us to identify the rpc instance
• Instance uptime
• Duration
• Request rate
• errors

Open questions?

• How can we prevent/discern corrupt/false data being sent?

[paulbellamy]

This is the metrics equivalent of: #34