[pdatautil] Add function for splitting log records while preserving scope and resource context

[djaglowski]

This type of function will be useful in the routing connector, where users often need to route telemetry based on contexts below the resource level. e.g. log records.

The trick here is to make copies of the resource and scope information only when needed.

A couple assumptions made in this implementation:

1. Empty scopes and resources are ignored, unless they became empty as a result of splitting log records. e.g. If all the log records from one scope are moved away, then that scope is deleted.
2. Rather than purely splitting an input into two outputs, it is more efficient to use the original input and create one new output. Much like actual pdata methods, a return parameter is used so that callers can provide their own preexisting target.
3. Moved resources and scopes are not merged into preexisting ones. (This isn't meant to be a batching function.)

[djaglowski]

I've converted this back to a draft for now, to consider how this could integrate with OTTL conditions, some of which would require construction of a full log context. We may need to pass resource and scope info into the split function as well.

[djaglowski]

Closing this in favor of a PR where similar functions are added to an internal package of the routing connector.