Add test and detailed error message for unsupported dependency case

[gbmarc1]

As per #183, if one package has multiple possible dependency versions, the dependency walker may fall into a trap where it discards the final "good" version.
The dependency walker requires quite a refactor to support this, and there are not tests to perform such a refactor in a confident manner.

Meanwhile, I propose this PR that at least details a bit better the problem so that users more easily investigate a work around.

[gbmarc1]

pre-commit.ci autofix

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
2 Code Smells

No Coverage information
0.0% Duplication

[radoering]

@dimbleby Thoughts on this? You don't happen to plan to solve this problem anytime soon? 😉

[dimbleby]

I have no plans to make fixes in this plugin.

As to this MR, I'm not convinced that it's actually doing anything very useful. Should we add a log saying "please contribute a fix" for every bug that we hit?

But I don't care to argue about it, if you're minded to merge then go for it.

[gbmarc1]

I can remove the contribute part of this PR. However, I think this more detailed message error identifies where the problem is (this plugging). I spent half a day investigating where the issue was.
When I encountered this error, I was looking if poetry's new minor version introduced a regression (1.32->1.4). I looked if I had an actual problem with my dependencies. The simple RuntimeError was too vague.
This PR make at least clear where the problem is, makes this plugin assumes owns its limitation, adds accountability, add cleanliness and a test!

I am opened to any suggestions to make the above objectives better.

[radoering]

I think it's good to merge:

• The error message is more detailed than before. (Dependency walk can fail for at least one other reason.)
• Mentioning poetry-plugin-export seems to make sense because many users are not aware of the difference.
• Probably, asking for contributions does not hurt since it shows we are aware of the issue. Of course, that might be a bit controverse and I agree that we shouldn't do it for any bug, but I'm fine with it in this case.