Reg2 update

[AntoinePassemiers]

Updated the regression_2 metric:

• Discarded target genes for which there is 0 putative TF according to the consensus.
• Clipped the coefficients of determination between 0 and 1 to avoid getting arbitrarily large negative values.

Improved the speed of single-omics methods:

• Added a parameter --M (number of iterations) in ENNET: now the default value is 100 instead of 5000.
• Added a parameter --nsplit in TIGRESS: now the default is 25 instead of 100. The nstepsLARS parameter is now set explicitly to prevent TIGRESS from setting it automatically to 1 (which produces errors).
• Reduced the number of genes in PIDC and scSGL by keeping only those with less than 75% of zeros (this threshold was set to 90% in the other algorithms). Also, made a docker container for scSGL so we don't need to clone the repo at runtime.

[janursa]

reviewed