Remove grouping trick

[dafeda]

Resolves: #200

[dafeda]

Ran with and without trick on a gaussian field with 1 million parameters, 50 observations and 25 realizations. The sigma=10 controls the smoothness of the field.
It runs faster without the trick.

    p = 1 mill, m = 50, N = 25, sigma = 10
    With trick:    1.8s, 1.7s, 1.9s (Mean)
    Without trick: 1.3s, 1.2s, 1.3s (Mean)

I also ran Drogon with and without the trick and it runs much faster without the trick
My conclusion for now is that the trick is faster when the field is very smooth which it does not seem to be in practice (Drogon).
So, I suggest we remove it.

[tommyod]

Since indices_of_row is now an int, this variable X_subset is now a 1D array. I am surprised that an error is not raised or that the code does not fail? Since X or a subset should be a 2D array.

Perhaps look at this quickly before merging. Also, perhaps rename

indices_of_row -> parameter_num
unique_row -> correlated_responses

or something like that. No other comments from me. Looks good.

[dafeda]

Have not looked into it deeply but I'm assuming broadcasting is helping us out in this: X_subset + cov_XY_subset @ T.
Thanks for the better variable names 👍

[dafeda]

I want to run a final test using Drogon before merging, but we have some problems with bleeding.