Random seed reproducibility

We should systematically initialize the seed for the optimizations. Sth like this:
```
seed = np.random.randint( int(1e6) )
np.random.seed(seed)
# Whatever optimizations we need to do
```
After this, the seed could also be store in the solution file.

This would guarantee total reproducibility and help us to get the correct graphs in a single attempt (once it has been found).