fix add_ssssaaaaaaaa fallback

[tthsqe12]

this was another of those "compiler bugs" :-)
On a serious note, the 64 bit windows box in spot #1 on @isuruf 's appveyor seems to NOT be using our lovely hand-crafted asm macros in longlong.h. This is an outrage as much of flint's speed on small operands derives from these macros.

[tthsqe12]

Here is an example of the problem caused by mp_bitcnt_t
https://github.com/wbhart/flint2/blob/trunk/fmpz_factor/factor_trial_range.c#L59
On 64bit windows with gcc (machine #3 on @isuruf 's appveyor), the mp_bitcnt_t is long unsigned int and the mp_limb_t is long long unsigned int

[tthsqe12]

machine #3 on @isuruf 's appveyor is completely broken.

[isuruf]

@tthsqe12, it worked in #548. Some PR must have broken it.

[isuruf]

euler_phi....
This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.
Exception (flint_mpz_pow_ui). Power too large

is an overflow issue.

[tthsqe12]

See #557 for why it is broken now. Once you have an undefined pointer cast like the one I cited above, anything can happen.

[tthsqe12]

I am strongly in favour of increasing our testing regimen.

[isuruf]

I am strongly in favour of increasing our testing regimen.

Me too. We can enable appveyor on this repo again. I can also add cygwin tests if needed. MinGW and MSVC tests are already there.