Lost wake-up in `concurrency_breadth_first_search`

[vil02]

Describe the bug

I run the tests with the thread sanitizer. The test_bfs takes forever to finish, which indicates that there might be some problem in concurrency_breadth_first_search. When I set num_threads to 1, the problem does not occur.

This seems to be the problematic code:

CXXGraph/include/CXXGraph/Graph/Algorithm/BestFirstSearch_impl.hpp

Lines 230 to 235 in 376d91b

	next_level_cond.notify_all();
	} else {
	// not to wait if last worker reachs last statement before notify
	// all or even further
	std::unique_lock<std::mutex> next_level_lock(next_level_mutex);
	next_level_cond.wait(next_level_lock, [&level, cur_level]() {

I think the problem is that the thread will not get notified by notify_all(), if it is not already in wait().
To Reproduce
Steps to reproduce the behavior:

1. Replace

   CXXGraph/CMakeLists.txt

   Lines 26 to 35 in 376d91b

   	if(SANITIZE)
   	add_compile_options(
   	-fsanitize=address
   	-fsanitize=leak
   	)
   	add_link_options(
   	-fsanitize=address
   	-fsanitize=leak
   	)
   	endif(SANITIZE)

   
   with

	if(SANITIZE)
		add_compile_options(
			-fsanitize=thread
		)
		add_link_options(
			-fsanitize=thread
		)
	endif(SANITIZE)

2. Run the build with -DSANITIZE=ON and run the tests (cf. you might need this).

Expected behavior
test_bfs takes reasonable amour of time.