event.success_count metric should use representative count

[lahsivjar]

event.success_count attribute is populated using ingest pipeline defined in apm-data plugin. The attribute currently is populated using 1 for success spans and 0 for failed spans, however, when we aggregate this metric, we use representative count (ref) which should be more apt for what we do with this metric.

The issue is about fixing the ingest pipeline to use representative count if defined, 1 otherwise in the ingest pipeline for cases when the event.outcome is success.

Related to: elastic/opentelemetry-lib#112

[raultorrecilla]

moving this to it106