Issue 27: Create truncnorm distribution

[ManuelAlvarezC]

This PR include the changes in order to add the truncnorm distribution from scipy:

• Created a new class Univariate.base.ScipyWrapper, it allows to integrate any subclass of scipy.rv_continous by just providing the name of the class and a mapping of the name of its methods, it's based on copulas.univariate.kde that already wraps a distribution from scipy.
• Implemented scalarize and vectorize decorators to help accept scalar values to vectorial methods and vice-versa.
• Add the class copulas.univariate.TruncNorm.
• Add a new implementation of GaussianKDE using the new ScipyWrapper class accepting and returning numpy.array, as requested on Make KDEUnivariate accept arrays as arguments. #85.
• Make vines and Trees use the new class instead of the old one.
• Add unittests for all the changes mentioned

Resolves #27, resolves #85, and resolves #90

[csala]

I think this could be a class level attribute, self.METHOD_NAMES. It also should be a tuple (immutable)

[ManuelAlvarezC]

Right, already changed.