pytest: port eip spec version checking to a pytest plugin

[danceratopz]

This PR ports the reference spec checking by implementing the functionality in a pytest plugin: spec_version_checker.

If the checks fail, they get reported as warnings in the pytest test summary.

See for example the following two runs that report warnings that:

1. pytest was unable to check the spec version, and,
2. the spec version is missing/notprovided, respectively, out of date.

There is no way from pytest's exit codes to tell whether warnings were raised. If we want to be able to check whether these warnings are present, via CI, for example, we'd have to run pytest as:

pytest -W error 

This won't run fillers though, as it's currently implemented, as the warning gets treated as an error in test setup. This will also treat all warnings in the test session as errors, with a bit more work, we could filter using our custom warnings using, for example,

pytest -W error::NoReferenceSpecDefined

In general, however, the output using -W error:: is very noisy as an error for every test is raised. The following is example output with -x (stop after first fail):

[marioevz]

Looks great! I like the plug-in approach definitely.

[danceratopz]

Thanks @marioevz. Perhaps we could try and improve this in the future by enabling the spec checks in a separate test(?). The idea being that it could be ran as a separate check. Currently, you have to execute all the fillers in order to check that the specs are up-to-date. I think this PR adds the existing functionality from tf to pytest, though.