[Feature Request]: Expose metrics for `rps`

[smuu]

Implementation ideas

Currently, we scale DA nodes based on memory usage.

@walldiss proposed scaling on rps, which would need to be exposed in order to scale on that metric.

cc. @Bidon15 @walldiss

[smuu]

It would be great to have a counter on the requests per type (like requested headers). This would enable the following:

• Scaling on this as resources are not sufficient enough
• We can compare requests per second with resource usage in our dashboards for debugging & monitoring

[walldiss]

There are multiple components that needs to be monitored to get an idea of rps / load:

• [x] shrex/eds server
  • metric:shrex_eds_server_total_responses
• [x] shrex/nd server
  • metric:shrex_nd_server_total_responses
• [x] header server
  • metric: hdr_p2p_exch_srvr_headers_served_counter
• [ ] ipld server
  • metric: not yet implemented, tracking issue: https://github.com/celestiaorg/celestia-node/issues/2454

[cristaloleg]

CC: @walldiss to revisit later