Consider adding support for `client.address` and `service.address` semantic conventions

[herrBez]

Preamble
According to the OpenTelemetry documentation, the client.address and service.address semantic conventions are considered stable . The *.address field, similarly to ECS can be both an IP-address and a hostname/domain name.

The APM Ingest pipelines apply the geoip processors to client.ip and service.ip to enrich the data with geographical information.

Issue

As of now, given that the fields are not recognized by apm-data, these fields are put into labels.client_address and labels.server_address and no enrichment is applied.

Desired Solution
It would be beneficial if the APM data processing could check whether client.address contains an IP address and, if so, populate the corresponding ECS field client.ip, so that the existing ingest pipelines populates the geo information

[gregkalapos]

My 2cents to this is that we should try to cover this functionality with the EDOT collector, but not in APM Server/MIS, because ultimately we'd want to move away from those in favor of EDOT.