Avoid multiple computation of ERI blocks for CVS calculations

[maxscheurer]

Right now the computation of ERI blocks is pretty optimal for importing full ADC ERI blocks, but for CVS blocks are computed multiple times. E.g. o1v1v1v1 triggers a computation of OVVV in the host program, then this block is flushed and o2v1v1v1 will again trigger the computation of this block, followed by a flush. It would be good to compute such blocks in a batch to avoid this behaviour.