fpu: Use plain 'int' rather than 'int_fast16_t' for exponents