Add "conflicts" constraints for each core implementing virtual VLNVs, so
no two cores implementing the same virtual VLNV may be selected
simultaneously.

This change should prevent more than a single implementing core from
being selected as the implementation for a virtual VLNV, and when there
is an explicit dependency in the tree, that one should be selected.

Signed-off-by: Alexander Williams <awill@opentitan.org>