Construct independant law from scenarios.

[leclere]

Allow user to define laws from scenarios of datas.

[frapac]

Currently, we estimate probability laws with Monte Carlo (we use different scenarios to build NoiseLaw). I think we should use Quantization instead.

http://www.quantize.maths-fi.com/

Or maybe;
https://github.com/EconForge/dolo/blob/master/dolo/numeric/discretization/discretization.py

[leclere]

As far as I know we currently ask for NoiseLaw from the user. To address this issue we can :
-- construct NoiseLaws as uniform on the given set of scenarios
-- find an optimal quantization for a given support size of xi_t. But this is pretty complex to implement.

[frapac]

I built this in an other repository, I will move the code ASAP