Introduced 0 (zero) to mp_is_square as a perfect square

[czurnieden]

Instead of just adding some documentation regarding issue #521 I decided to fix that bug.

• changed definition of 0 (zero) from not-a-square to is-a-square.
• test added to check 0 (zero)
• additional test added to check negative numbers

Does not seem to conflict with LibTomCrypt but that needs a proper check, I just grep'd for is_square.

Changes API and hence might conflict with other software that uses LTM but on the other side: it was a bug that gets fixed with this patch.

[sjaeckel]

Changes API and hence might conflict with other software that uses LTM but on the other side: it was a bug that gets fixed with this patch.

How does this change the API? FMU it only fixes the bug.

Does not seem to conflict with LibTomCrypt but that needs a proper check

LTC doesn't even have a direct mapping of mp_is_square() into the MPI provider, so it doesn't matter.

The single LTM-internal consumer is mp_prime_is_prime() and there the behavior doesn't change since 0 is even and therefor we return before mp_is_square() would even be called.