More granularity to the quickwit_http_requests_total metric

Is your feature request related to a problem? Please describe.
When looking at a cluster or a node, it is very useful to correlate "behavior metrics" (e.g resource usage) with "usage metrics" (number of requests, ingested bytes). As the HTTP API is one of the core user interfaces, it would be good to have a more accurate view on the kind of endpoints that are used. For instance, we want to be able to filter out healthcheck calls. It is also very useful to detect unexpected client activities (e.g search queries sent to indexers).

Describe the solution you'd like
Add an "endpoint" label to quickwit_http_requests_total

Describe alternatives you've considered
We could also group endpoints to decrease cardinality, but the information loss is probably not worth the benefit, especially when taking into account that a node is usually specialized (e.g search) and only receives requests to a small subset of the API.

Additional context
Add any other context or information about the feature request here.

Start by locating the quickwit_http_requests_total metric definition and the HTTP API request instrumentation. Trace how requests are associated with endpoints, then verify that the metric exposes an endpoint label and that healthcheck and other API requests can be distinguished in the exported metrics.