Remove/allocate unused numpy arrays to avoid memory leaks

As found by @simon.maenaut: if you pass a numpy array to a cython function and it is not used within the function, it does not really get cleaned up and can cause memory leaks, see: https://github.com/cython/cython/issues/3046

This MR:

1. Removes unused variables [**Done**];

2. Always allocates the ones that are used only in certain cases. Yet to do:

   Specifically inside the likelihood:

  time_array_raw, ACF, inverse_Covariance, cholesky