math/big: modInverse doesn't require a prime modulus

[randall77]

The spec of math/big.modInverse requires the modulus to be prime.  But the inverse is
well defined as long as the number being inverted and the modulus are relatively prime. 
Perhaps more importantly, the current code seems to work for all of those cases.

Should we relax the spec to include these cases?
Any reason why the current code wouldn't always work with this relaxation?

[agl]

Comment 1:

There's no problem with relaxing that, but the prime modulus case was the only one that
I cared about and tested and thus the only one that I wanted to commit to.

[gopherbot]

Comment 2:

CL https://golang.org/cl/155010043 mentions this issue.

[randall77]

Comment 3:

This issue was closed by revision 96d1e4a.

Status changed to Fixed.