Handle cyclical requires

[dcbaker]

By either ignoring them as an invalid solution, or returning an appropriate error when there is no valid solution

[bretbrownjr]

We should allow modelling of cyclical graphs. It's a real situation that becomes an adoption hurdle when it happens. In particular, codebases that are messy shouldn't be forced to completely get their acts together before adopting CPS or any tools that require well formed CPS in the distant future.

It's probably OK to make support for cycles an explicitly opt-in feature though.

[dcbaker]

I looked at what pkgconf did again, I thought that it was aborting on cycles, but it breaks the cycle. I think it would be worth doing what pkgconf does. But we need to do anything other than what we do now, which is recurse until the stack overflows

[bretbrownjr]

It's a leetcode not-a-big-deal problem to walk a cyclical graph, so I'm sure we can manage. Someone write this up like it's leetcode, mark it help-wanted, and I'm sure someone will pick it up in due time.

I don't know if we need it for an MVP, to be clear.