Manage automatically similar series base

[gcabillic]

When series are similar, we should add a way to reduce automatically the number of samples involved in the regression computation. We could adjust automatically the base to keep a maximum of the last 50 or 25 builds in that case?

The regression computation will not be impacted and the computation time of the regression analysis will be better.