New attribute for "worker.id"

[jpkrohling]

Some frameworks and languages are typically executed by an application server such as unicorn (Ruby). In those cases, the same application is being executed multiple times by the same parent process (and potentially the same PID), but they represent different instances of the same service.

Therefore, I suggest adding a new semantic convention to the process runtime resource, recording the worker.id for that instance. This can later be used as part of the algorithm that generates the service.instance.id semconv.

Originally posted by @jpkrohling in #312 (comment)