Add handling of MP_ITER error-code to mp_error_to_string()

[nijtmans]

mp_error_to_string() doesn't handle MP_ITER currently, I think it should.

[minad]

While you are on it - what about replacing the function with a simple switch statement? With modern compilers you can enable coverage checking then.

Edit: Ok we are not using enums :(

[sjaeckel]

Edit: Ok we are not using enums :(

but we should

Edit: but not in this PR

[minad]

No, not here, it would also give some kind of compatibility issues. This time in the public API, about which I am much more hesitant. My should was more meant as - this is how it should be done in a modern c style.

[nijtmans]

I agree with the comments: It should have been an enum, but that can cause compatibility issues in the public API, so doing that now is a bad idea!

[minad]

Haha, @sjaeckel you wrote we should. I only thought it :)

[czurnieden]

mp_error_to_string() doesn't handle MP_ITER currently, I think it should.

Yepp, forgot it, my bad.
But at least it is documented in bn.tex ;-)

And it is not really an iteration that doesn't converge in a timely manner but the list of available candidates for the coefficient a ended (significant only with MP_8BIT, although the probability is sufficiently low), so MP_ITER is another proof that I am really bad at naming things.