Bug fixes and new sampling method

[vwcruzeiro]

In this PR I am modifying the following:

1. Correct a mistake in the expression for the distortions of each normal mode that was effectively making the temperature 3 times larger than the temperature inputted by the user. In the equation 5 of the ANI-1 paper, instead of 3Na I am now placing Nf=3Na-6. By doing this, we don't give to the NMS energy from the translational and rotational modes. Also, if the user arbitrarily removes modes from the input files, by doing this we don't give more energy than we should.
2. Fix getCharge inside lib/nmstools.py
3. Modify the algorithm for the random number generator inside the NMS code at lib/nmstools.py
4. Adds a new type of sampling that I called "structural sampling". What this does is the following: we can place a bunch of xyz files in a folder and randomly feed these structures to the active learner. This is a nice feature because we can use external software to generate structures and then just feed it to ANI's AL.