Hi there,

I know it's not A HUGE PROBLEM, the library does the job fine (thanks for your work!), but I noticed that the names of the metrics the distribution creates are not in line with recommendations from Prometheus docs. For any metric name given by the user, this library will append three suffixes: bucket, sum, and count. Let's consider something like handling an HTTP request.

If we go with http_request_duration_seconds, then http_request_duration_seconds_count looks weird.
If we go with http_request then count looks fine, but sum is off, and we don't follow the recommendation to include the unit in the bucket metric name.

So I tried to see if something can be done about it. What I came up with is allowing the user to include the suffixes to use in the reporter options. The metric definitions then looks like this:

distribution_reporter_options = [
  bucket_suffix: "duration_seconds",
  sum_suffix: "duration_seconds_total",
  count_suffix: "total",
  buckets: [
    0.005,
    0.010,
    # ...
  ]
]

Telemetry.Metrics.distribution(
  "http.request",
  event_name: "http.request.stop",
  measurement: :duration,
  unit: {:native, :second},
  description: "Total HTTP request handling time",
  reporter_options: distribution_reporter_options
),

Which results in this output:

http_request_duration_seconds{le="+Inf"} 111
http_request_duration_seconds_total 0.010809296000000003
http_request_total 111

The PR only includes the bare minimum of code to make it work (no validation, no docs), because I just wanted to ask if this is something you'd be interested in.

Best regards