* Locally ASAN wasn't complaining.

* Add MultiProcessingTimer struct
* Replace raw timespec and double counters with MultiProcessingTimer(s)

* Introduce a small public domain timespec handling library
* Renamed MultiProcessingTask members _now and _seconds to _interval_start and _interval_stop
* Added interval_elapsed member
* Change elapsed member from type int to double
* mp_pool_free no longer tries to free semaphores with empty names
* semaphore_init uses the correct default value of 1 instead of 2. The bug was related to calling semaphore_wait before the child() process started and semaphore_post after the parent exited. Now we post before the child to prevent a deadlock
* Replace sleep with usleep in mp_pool_join. Set to 100ms.
* Replace time() with clock_gettime() (helper functions created to prevent mistakes)
* Stop recording time when the process ends. This fixes a bug where a process ends and up to 3 different elapsed times were reported to the user.
* Progress output is now always available, not only when pid > 0
* Implement seconds_to_human_readable and hook it up to progress reporting calls. Breaking down thousands of seconds in my head after a long run was mentally exhausting.
* Cleaned up some if-statements; removed else-if for clarity
* Implemented a global timeout for pool tasks
* Add register_semaphore() and semaphore_handle_exit() to aid with clean up. On Darwin a dangling shared memory file will lead to unexpected failures. These are destroyed via atexit().

* Add argument: --task-timeout=1[s,m,h]
* Timed out tasks are SIGKILL'd
* If killing a task fails, the entire program ends