Refactor metrics to use alternate metrics crate

[adam-cattermole]

Changes

I've removed the original prometheus crate and added metrics, and prometheus metrics exporter. These support using the metrics API, and the exporter allows us to get a handle to be able to still use actix to export the metrics at /metrics.

• Metrics are configured with the help message via the describe_* methods on init
• They are no longer stored in the struct and used from the macro directly (due to the macro supporting labels but no method on the metrics types)
• The prometheus_metrics struct is still passed to envoy_rls/server and to the http_api/server to retain the config limit_name_in_labels for writing the metrics - not 100% sold on this solution though
• The prom metrics struct now retains a handle to the global prometheus recorder, as this can only be set once for the lifetime of the application. The tests share a single lazy static handle
• I've provided an example for adding a datastore_partitioned metric to the limitador library

Verification

Checkout this branch and build server:

cargo build --bin limitador-server

Run the server:

./target/debug/limitador-server -vvv --grpc-reflection-service --limit-name-in-labels limitador-server/sandbox/limits.yaml redis_cached redis://127.0.0.1

Check metrics endpoint:

curl localhost:8080/metrics
# HELP limitador_up Limitador is running
# TYPE limitador_up gauge
limitador_up 1

TODO

• Fix tests, unable to configure the prometheus exporter

[alexsnaps]

🥳