Add checkpointing capabilities to L2L

[sdiazpier]

Enable the possibility that a user restarts an optimization run using an existing trajectory and python script.

• [x] Make this an option that can be enabled from the python script
• [x] Load the trajectory that the user provides as last checkpoint
• [ ] Check that the optimizee is available to be executed
• [x] Make sure the generation number matches the generation of the last trajectory
• [x] Make sure all data structures used by the optimizer are correctly initialized
• [ ] Test that all optimizers can use this functionality without problems