monitoring: Record event type metrics from client in pre batch processors

Since EventCounter is after TBS and aggregations, `apm-server.processor.span.transformations` and other exposed metrics are post-aggregation (i.e. metric counts would include aggregated metrics) and post-TBS (i.e. it only reports sampled span counts).

We should add an EventCount in the pre batch processors to report actual event counts from clients.

- naming is TBD, whether it should reuse `apm-server.processor.span.transformations` etc and move the existing one to use another metric name.
- update metric mappings in integrations, metricbeat and stack monitoring mappings.