Use protected_int*_adjust[_fetch]() correctly

At least on Linux/GCC, std-atomic version of xpdev/protected_int API was only
used when building .cpp files and the _adjust() functions returned a different
value from the .c files since they used the mutex versions of these functions/
macros. Use the _adjust_fetch() versions of the API when the return value is
used (and assumed to be the *new* / adjusted value of the integer).

It's possible for other tool-chains or target platforms, the .c usage was
wrong as well (assuming the *new* integer value was returned rather than the
old), if they built with std-atomic support for C.