refactor(pytest): make `spec_version_checker` plugin independent of the `filler` plugin

[danceratopz]

Currently, the spec_version_checker collects every "normal" python test case and then detects all the test nodes that contain "eip" in their name. For these tests, a new test eip_version_check test is dynamically added for every test module.

As the spec_version_checker collects the "normal" tests, it must be be able to execute all the fixtures that they use (from fill), which isn't so bad, it's just a bit noisy/ugly. It'd be cleaner to simply collect the python test modules and then generate a test for each module, instead of depending on full regular test case collection.