Correct bracket for exponential

[wreise]

This PR aims to fix two issues noticed after #755 :

• Incorrect bracket in exponential makes tests pass almost always.
• Many lines are marked as tested only because they are covered by src/python/example/diagram_vectorizations_distances_kernels.py.

[wreise]

After fixing @mglisse 's comments, now the tests crash (at least locally): the approximation seems to be not that good. I will have a closer look.

[wreise]

The approximation with RBFSampler is consistent if
PersistenceWeightedGaussianKernel(bandwidth, weight_fct)(dgm1, dgm2) is equal to the expectation of PersistenceWeightedGaussianKernel(bandwidth, weight_fct, kernel_approx=RBFSampler(gamma=gamma))(dgm1, dgm2). It is actually the case only when gamma = 0.5*1./(bandwidth**2).
Would it be useful to mention that somewhere (or to even restrict the API to force consistency)?

[MathieuCarriere]

Sorry for taking so long to answer. I am not so sure about restricting the API, given that RBFSampler is also used to approximate other kernels (I think). But it would definitely be good to add a comment about this in the doc of PersistenceWeightedGaussianKernel!