Provide an exception message for BadRef

[Jacoby6000]

I ran in to this while translating some non-trivial json schema specs. Without this change, bad refs evaluate to null when getMessage is called.

Also updated the trait definition such that all traits must provide an error message or else compilation will fail.

[CLAassistant]

All committers have signed the CLA.

[lewisjkl]

Hey @Jacoby6000 thanks for contributing! I like the change, do you think you could also add a test case that shows what this fixes so we can avoid future regressions?

[Jacoby6000]

I went ahead and added the test suite. In hindsight I'm not so sure this is actually testing anything, given that the compiler now enforces that messages must not be null

[lewisjkl]

I went ahead and added the test suite. In hindsight I'm not so sure this is actually testing anything, given that the compiler now enforces that messages must not be null

Right, I was more-so thinking of a test where the input was a reduced version of whatever you ran that caused the error to pop up originally. That way we have a test that, without your fix, would fail.

[Jacoby6000]

a test where the input was a reduced version of whatever you ran that caused the error to pop up originally. That way we have a test that, without your fix, would fail.

Ah, that is addressed in #273

This PR doesn't really fix anything. It just improves the developer experience when they come across a BadRef