Fix the atomicity of cy_atomic_int

[ptrrsn]

When building sage, meson.build incorrectly sets all CYSIGNALS_* macros to 0 because of lacks of main function, which leads to a non-atomic declaration of cy_atomic_int.

Moreover, the cy_atomic_int typedefs in struct_signals.h was structured in a way that when it is compiled by a C++ compiler in an environment with a C compiler that supports _Atomic, it would never use std::atomic, which is incorrect and might cause compilation problem when the C++ compiler does not support _Atomic. This problem was hidden by the incorrect configuration of meson.build above but got surfaced after the meson.build fix.

As a context: LLewark/khoca#3

[orlitzky]

int main(void) please, for consistency across the various C standards

[ptrrsn]

Thank you for the review. I changed int main() to int main(void) for the C atomicity checks.

[dimpase]

I'll draft a new release later today

[ptrrsn]

Thank you for the reviews and merge.

[dimpase]

@ptrrsn I've put 1.12.4 with this fix on PyPI (so pip install cysignals should get it)

[ptrrsn]

Thank you, I can see it now. cc @soehms

[soehms]

Thank you, I can see it now. cc @soehms

Thank you! For the record: Before applying 1.12.4 I got at least 2 failures with ./sage -t --global-iterations 20 src/sage/knots/link.py on each run. After upgrading cysignals, reinstalling khoca and cypari2 an rebuilding Sage I don't get failures any more on several runs.