optimize primary peak-motifs discovered matrices rather than clustered matrices

[jvanheld]

So far we ran optimize-matrices-GA on matrices resulting from peak-motifs | matrix-clustering | trimming.

With matrix optimization, it is not sure that we need the clustering step. Starting from the peak-motifs discovered motifs would give more traceability on the source of the motifs retained at the end of the optimisation, and possibly tell us if some algorithms (oligo-analysis, dyad-analysis, position-analysis) give generally better results, or if this depends on the TF/dataset.