Support data pickling

[rstrudel]

Hello,

First, thanks for adding the support of model pickling I mentionned in #957 ! I noticed that the same error occurs when trying to pickle data objects. If it is possible, it would be nice to have both model and data instances picklable.

Here is a short snippet which produces an error:

import pickle as pkl
import pinocchio as pin

model = pin.Model()
data = model.createData()
pkl.dump(data, open('data.pkl', 'wb'))

Cheers,

Robin

[jcarpent]

Do you really need the pickling of data? In which context?

[rstrudel]

For multiprocessing purposes I believe not being able to pickle an object is problematic. If you want to compute a function which output a Data instance (using joblib or dask for example), it will output a pickling error.

Along the same lines you cannot pass a RobotWrapper instance to a function you want to parallelize because pickling is required.

Another example is when you want to snapshot a running experiment. Pickling the current state of the robot allows you to check and debug later on.

[jcarpent]

Solved by #998