CI: Another attempt to fix windows ci..

[M2-TE]

Seeing how it failed again, I dug a bit deeper into what could cause it.

Flushing the std::ofstream seems to only guarantee synchronization for the current application. The OS has to decide what to actually do, where Windows seems to have some trouble, as we spawn a separate process with std::system.

From what I found, there are two solutions, one being a hacky one that someone suggested on SO:
https://stackoverflow.com/a/9667875
which essentially just opens a new stream, which seems to force windows to actually finalize the write to disk.

There might be a better solution though, looking at the spec of std::system itself:
https://en.cppreference.com/w/cpp/utility/program/system

An explicit flush of std::cout is also necessary before a call to std::system, if the spawned process performs any screen I/O.

Talk about fine print..
Anyways, I've reverted the ofs to what it was before, but added a std::cout.flush() right before we do the clangd-formatting step. If this doesn't work, the reopening of the ofs in append mode or a separate std::ifstream seems like a good solution.

[asuessenbach]

Wow! That seems to be a great catch.

Thanks a lot for investigating on that issue. And it seems, you've found a resolution!