Add transcript length annotation to gene data

[dakota-hawkins]

allows on the go tpm conversion.

normalize and batch correct counts in scran -> tpm when needed (i.e monocle).