Tests that (almost) always fail

[cboulay]

• [ ] MINGW - internal loguruthreadnames skip
• [ ] MacOS - internal reuseport on ipv6 multicast - skip
• [ ] pushpull benchmarks in MSVC win often segfault

The first 2 don't test LSL code directly so I have excluded the problematic config on the problematic target platform.

[cboulay]

I managed to reproduce a segmentation fault from the pushpull benchmark test locally. It happens here:

https://github.com/sccn/liblsl/blob/7e61a2ef7a25a54b54968fb5c8776054f49b028e/src/sample.cpp#L81

It happens at the very end of the test case, as the outlet is going out of scope, triggering its cleanup:

lsl::stream_outlet::~stream_outlet() > lsl_destroy_outlet(...) > ~stream_outlet_impl() > lsl::factory::~factory()

During ~factory, this line https://github.com/sccn/liblsl/blob/7e61a2ef7a25a54b54968fb5c8776054f49b028e/src/sample.cpp#L470

cur is already nonsense.

[cboulay]

220 might be helpful: "bad_weak_ptr"