Suppress version constraint token scanner errors

[wagoodman]

The text/scanner is used for constraint expression tokenization. If no error handler is provided to the scanner then errors are reported to stdout by default. This PR adds a handler which ignores all scanner tokenization errors, which is safe under the current context (see the added comments for details).

Closes #246

[luhring]

Weird af!

[alfredodeza]

I'm guessing there is no easy way to test this one out right? Otherwise, I would love to catch this with a test, brownie points if it can catch other unexpected output as well as the one this fixes.

[wagoodman]

@alfredodeza I thought about what kind of test I would write and how to write it. It can't easily be done in unit testing without affecting other tests, so that's not an option. Adding a new CLI-level test to account for it was something I considered, but felt it to be a little heavy handed for testing if the stdlib scanner configuration was correct (make certain this specific error text does not appear).

I think we can add a more generic test like you described last (check that output matches the given output, fail if anything else is given), but can be added in a follow up PR as a CLI test to verify grype scan output for multiple test-fixtures. That test would add a lot of value outside of this specific behavior being fixed on this PR.