bug in __gmp_replacement_vsnprintf

[Paul Zimmermann]

       Hi,

[for the record, this issue was originally reported on the MPFR list:
https://sympa.inria.fr/sympa/arc/mpfr/2022-10/msg00001.html]

Originally, it appeared only under Windows with the clang compiler,
and using MPIR, but I can reproduce it under Linux with GMP 6.2.1:

1) configure GMP
2) uncomment the #define HAVE_VSNPRINTF 1 line in config.h
3) build GMP
4) run the MPFR tsprintf test file with the built GMP

The issue is because __gmp_replacement_vsnprintf does not deal with %a not %A.
Then when calling gmp_printf ("%a", -1.25) for example, we get total_width=3
initially, we jump to the 'default' case, where the ASSERT(0) does nothing
in production code, and we go to next, where width=0 and prec=6, thus
total_width is increased to 9. But we also have len=9 because
buf='-0x1.4p+0'. Then the assertion ASSERT_ALWAYS (len < total_width) fails.

Paul