Thoughts on improvements to speed efficiency

Notes on speed/efficiency:

-- even if dimReduce="none", every call to clusterSingle calls .transData to do the transformation of the data. Might be slow. Has call x<-transFun(x) (copies over)

-- How much do we loose by clusterSingle making clusterExperiment object repeatedly (so multiple assay). Perhaps should give option in clusterSingle to return just vector of cluster identifications and have clusterMany put it together into clusterExperiment object.

-- Is there reduced memory if not make summarized Experiment object first to clusterMany? (i.e. use matrix input)