Support monotonic operation for counter/sum

[xuan-cao-swi]

The counter should be adjustable for monotonic operations so that, when enabled, it prevents any negative values.

Related discussion: #1778

Related docs:

• https://opentelemetry.io/docs/specs/otel/metrics/data-model/#opentelemetry-protocol-data-model
• https://opentelemetry.io/docs/specs/otel/metrics/data-model/#sums
• https://opentelemetry.io/docs/specs/otel/metrics/data-model/#timeseries-model

[garoazinha]

I also couldn't export metrics to prometheus and could fix it by changing the way the exporting is done. According to the specs non_monotonic type sum is based on the instrument. Doing this:

case instrument_type      
 when :counter
  Opentelemetry::Proto::Metrics::V1::Metric.new(
    name: metrics.name,
    description: metrics.description,
    unit: metrics.unit,
    sum: Opentelemetry::Proto::Metrics::V1::Sum.new(
      aggregation_temporality: as_otlp_aggregation_temporality(metrics.aggregation_temporality),
      data_points: metrics.data_points.map do |ndp|
        number_data_point(ndp)
      end,
      is_monotonic: true
     )
   )
end

on the metrics otlp exporter already solved the issue for me. The is_monotonic attribute should be tied to the Sum class though, as it is in other implementations. The parameter is usually set when initializing the sum aggregation object, in:
python impl
js impl
go impl

I am trying to do something similar but unsure if this is the way to to go.