really the square of the return value.

BN_mod_exp_mont does not work properly yet if modulus m
is negative (we want computations to be carried out
modulo |m|).

(just in cases someone uses a negative modulus)

Fix BN_mod_exp_simple for a==0 (mod m).
Skip useless round in BN_mod_sqrt (1 is always a square, no need
to test BN_kronecker for it).

Also add the next file in advance so that I can't forget this one :-)