Initial metrics support

[CHr15F0x]

This PR is the second attempt at adding metrics to pathfinder (feature request issue, first attempt),

• It's based on the metrics and metrics-exporter-prometheus crates.
• The default prometheus exporter is not used, instead the payload is rendered in the /metrics route.
• Only the basic total call counters were added for RPC methods to keep the PR small.
• RecorderGuard should be used when installing a per-test recorder, which takes care of recorder removal and avoids races between tests (metrics forces a global recorder instance per binary).
• At the moment I'm using format!() to build the counter labels but this can be sped up ~5x if using a BTreeMap<&'static str, &'static str> for method name vs counter name lookup instead (benchmark is disposable so not included in the PR). I haven't added this yet as I assumed that this extra heap allocation is negligible in terms of performance when taking the DB operations into account.

[koivunej]

At least Cc: #258

[Mirko-von-Leipzig]

No major complaints :)

Is there anything we can use to see what this produces externally?

[koivunej]

I would hope that the RPC changes could be split off, done in a follow-up. I failed to see those were just really strangely diffed whitespace changes. Otherwise this looks good, would:

• consider changing adding in a metrics prefix right away since this will become part of the "stable api" and we don't want to go back and forth with it often
• would consider not initializing any recorders from main when monitoring addr is not specified, or at least commenting why it needs to be always added

I don't really know what would be a good way to handle the format! away. Cheapest at runtime might be https://docs.rs/phf/latest/phf/ but hope you kept that BTreeMap bench :) But yeah, it's followup stuff.

[CHr15F0x]

Once labels are used to differentiate between counters for different method calls there is no need for format!. I did a quick benchmark and method_name_as_non_static_str.to_owned() in on_call() turns out to be on par with the fastest of lookup alternatives (these include phf::[Ordered_]Set which surprisingly is not the fastest).

[CHr15F0x]

@Mirko-von-Leipzig

Assuming monitoring is on port 9000 you can just:
curl localhost:9000/metrics
And you should see

# TYPE rpc_method_calls_total counter
rpc_method_calls_total{method="starknet_getTransactionByBlockIdAndIndex"} 0
rpc_method_calls_total{method="starknet_getBlockTransactionCount"} 0
rpc_method_calls_total{method="starknet_getTransactionByHash"} 0
rpc_method_calls_total{method="starknet_getEvents"} 0
rpc_method_calls_total{method="starknet_getClassHashAt"} 0
rpc_method_calls_total{method="starknet_addDeployTransaction"} 0
rpc_method_calls_total{method="starknet_getStateUpdate"} 1
rpc_method_calls_total{method="starknet_blockNumber"} 0
rpc_method_calls_total{method="starknet_chainId"} 0
rpc_method_calls_total{method="starknet_blockHashAndNumber"} 0
rpc_method_calls_total{method="starknet_addInvokeTransaction"} 0
rpc_method_calls_total{method="starknet_getClassAt"} 0
rpc_method_calls_total{method="starknet_getBlockWithTxHashes"} 0
rpc_method_calls_total{method="starknet_syncing"} 0
rpc_method_calls_total{method="starknet_pendingTransactions"} 0
rpc_method_calls_total{method="starknet_call"} 0
rpc_method_calls_total{method="starknet_getStorageAt"} 0
rpc_method_calls_total{method="starknet_getClass"} 0
rpc_method_calls_total{method="starknet_getBlockWithTxs"} 0
rpc_method_calls_total{method="starknet_addDeclareTransaction"} 0
rpc_method_calls_total{method="starknet_getNonce"} 0
rpc_method_calls_total{method="starknet_estimateFee"} 0
rpc_method_calls_total{method="starknet_getTransactionReceipt"} 0

Or better download and run prometheus with a small addition to the config file (prometheus.yml) in scrape_configs, for example:

scrape_configs:
  # The job name is added as a label `job=<job_name>` to any timeseries scraped from this config.
  - job_name:       'pathfinder'

    # Override the global default and scrape targets from this job every 5 seconds.
    scrape_interval: 5s

    static_configs:
      - targets: ['localhost:9000']