Generic `Connection.process()`?

While the python client provides a lot of helpers/convenience functions/methods for standard openEO processes, there is sometimes need for generally "starting" a process graph from some non/less-standard openEO process (e.g. user defined processes or remote process definitions).
For that, we currently have Connection.datacube_from_process() (with some variants).

This helper dates back from 2020 (9db6452a8aea9ebcaa3e), when DataCube was roughly the only process graph abstraction/wrapper in the client. Since then however quite some other variants were added, like vectorcube, mlmodel, stac_resource, etc.
This makes datacube_from_process() insufficient:

• its name refers to datacube, while it should be about other types too
• it should return the correct wrapper based on the openEO process return type

I'm considering to add Connection.process() to handle this, but a bit hesitant, because "process" as name is quite generic, and can lead to confusion: does it refer to the noun process or the verb? Does calling it trigger processing (e.g. like there is already Connection.execute()) or is it just about building a process graph?

related issues:

• https://github.com/Open-EO/openeo-python-client/issues/844
• https://github.com/Open-EO/openeo-geopyspark-driver/issues/1476#issuecomment-3717996742

Start by reading Connection.datacube_from_process() and its variants, then review related issues 844 and the linked driver discussion. The work is not ready to implement until the API naming and wrapper-selection behavior are agreed; done should cover nonstandard process graphs and return the appropriate wrapper for each openEO process return type.