Change lock mechanism

Our current lock mechanism relies on a library that doesn't work properly if three or more instances of the experiment are running. ( mainly due to the shared filesystem )

• The lock is done properly on the first run,
• The lock is released if another run or create on the same experiment is triggered, although the attempt is blocked if the first instance didn't finish.
• The lock is done again, but the instance is never blocked.

The idea is to change this library to a custom file-based lock.