Return an error instead of panicking when encountering unsupported registers during unwind

[m-mueller678]

This fixes #55.

I have changed Frame::unwind to return a unit error type instead of a gimli error. The error information is currently discarded at all call sites of Frame::unwind. I have considered two alternatives:

• introduce a custom error type than can be either a gimli error or an unsupported register error. Given that the information is not used anyway, this feels like needless complexity.
• reuse the UnsupportedRegister variant of gimli's error type. This feels misleading, especially in the context of gimli's documentation:

  Registers larger than u16 are not supported.

This fixes the problem I had encountered in #55. However, I am unsure of what testing procedures unwind uses. I have only run cargo check and cargo test. The test output looks the same as before my changes.

[nbdd0121]

I suppose we might also be able to skip over any unsupported registers when unwinding (for phase 1 / stacktracing only, obviously we can't actually install contexts in such case).

[m-mueller678]

Could that not break unwinding of future frames? I am not very familiar with unwinding, so I do not know how things could break when resuming with wrong register values