Add custom flush function feature

[donghoonpark]

Like in #105, a custom flush function can help reduce the blocking time of the read API, especially when using an intermediate buffer with DMA. I’ll also update my embedded example if this pull request is accepted.

This PR also includes the following minor changes:

• __suseconds_t is not available on Apple arm64, so I added an alias for it.
• Added a launch.json configuration to debug the tests directly in VS Code using the cppdbg debugger.
• Edited tests/nanomodbus_test.c to pass clang-format ci step

I've already ran tests and checked all passed.

• unittest & build test : https://github.com/donghoonpark/nanoMODBUS/actions/runs/19188610342
• clang-format : https://github.com/donghoonpark/nanoMODBUS/actions/runs/19188665250

[taraskornuta]

I did a similar thing, but I try to keep original code
It makes nmbs->platform.flush an optional

static void flush(nmbs_t* nmbs) {
    if (nmbs->platform.flush) {
        nmbs->platform.flush(nmbs->platform.arg);
    } else {
        nmbs->platform.read(nmbs->msg.buf, sizeof(nmbs->msg.buf), 0, nmbs->platform.arg);
    }
}

[donghoonpark]

@debevv Could you please check this PR?

[debevv]

Hi, sorry for the absence.
@donghoonpark why does it return int now? Maybe to communicate the bytes discarded? If yes, are we sure all users will be able to get this information from the underlying system?

The rest seems all ok

[donghoonpark]

My intention was exactly same as you mentioned in embedded system.
When it comes to os including posix, windows do not give the information about the count of dropped bytes but gives success/fail flags. As it also a int type, I thought that it would be better to keep the return type as int.
Just let me know the way you preferred and I'll revise it. Or you can change the type to void before merge. :)