Reduce number of n loops for the finite time methods

[riberaborrell]

After adding backward_transitions(), forward_committor() and backward_committor() methods, the methods in pytpt/finite.py which have a loop through the time n are the following (UPDATE):

• density()
• backward_transitions()
• forward_committor()
• backward_committor()
• committors()
• reac_density()
• reac_current()
• current_density()

We would however keep using the method 'committors' to compute in one loop over n

• the density,
• the backward transition matrix and
• the committors

and we could add a method called 'reac_density_and_current' to compute in one loop over n

• the reactive density, the normalization factor and the normalized reactive density
• the reactive current and the currentcurrent_density.

Altogether we would just need 2 loops to compute all the statistics needed.

[LuzieH]

@eborrell I think depending on how much the speed up of using committors compared to backward_transitions/backward_committor and forward_committor is, I would just delete committors (because the other methods are clearer and nicer to use).
Do you know how the computation time compares?

[riberaborrell]

Depends on the size of N and the size of S. I adapt the test so we can compare now.

• For example, if S is small (case of the small network), there is not that much difference even for n >> 1.
  $ pytest tests/test_finite.py::TestFinite::test_separate_committors --small-network -N 10000
  $ pytest tests/test_finite.py::TestFinite::test_together_committors --small-network -N 10000

• For bigger S, the methods separated are slower
  $ pytest tests/test_finite.py::TestFinite::test_separate_committors -S 1000 -N 1000
  $ pytest tests/test_finite.py::TestFinite::test_together_committors -S 1000 -N 1000

I would keep both.

[riberaborrell]

update: committors() method removed.