A generalized multi-task IBSS framework incorporating SuSiE, mvSuSiE, SuSiE RSS and fSuSiE

[gaow]

If #3 is done, we can also actually incorporate susie_rss into it so we can also naturally integrate GWAS data. I think it might even be possible to extend this to taking mvSuSiE learners and mix with functional regression, RSS and regular SuSiE learners. This should be a general framework that naturally fits into our current engineering structure (susieR 2.0 backbone) since mvSuSiE, SuSiE and fSuSiE are now unified under this structure.

[gaow]

One conceptual challenge might be how to handle the residual variance if correlated. The other conceptual question is, how much benefit does it has over just run various different SuSiE models until convergence and come up with a way to "colocalize" these multiple models ( @xueweic ).

To answer my own question, I think a genome-wide version of this would actually justify this joint model because of shared genome-wide parameters --- this would actually connect to the susieAnn structure ...