Add plugin forward model step validations based on external config files

[yngve-sk]

(From discussing #8847)

Some forward models in ERT like eclipse sometimes depend on external config files, that may be generated by previous forward model steps and thus only be available in the runpath. In Everest-models something similar is done, where the user gives yaml files that configure how certain forward models run. These yaml files are also validated by Everest, but here Everest mocks up a runpath. It is not really clear how this should be resolved for ERT, but if it is resolved here first, it should be more clean to port the Everest forward model validation over to this functionality also. We assume that pre-experiment validation will only done if the config file is given up-front. If it depends on {{ realization }} etc, we can maybe use whatever is done w/ install_data in Everest (Note: not sure if this is completely correct)

Proposed steps:

1. Set up an ERT example with a plugin forward model is configured by an external config file (***that is not generated by a preceding step which dumps it into the runpath).
2. Write up some code that does the validation, given the path to a config file, add it to the PluginForwardModelStep ABC, add an invocation to it at an appropriate place in the ERT flow of execution and add tests for (1) failed validation of the config file, (2) successful validation of config file.

*** IF it is generated by a preceding step we can't really do a pre-experiment validation on it anyway, or can we?

Possibly related: #9036