radv: avoid bottlenecking on sequential sparse buffer binds
it's more costly to submit individual sparse buffer binds than to merge them and submit bigger binds, so try to pre-compare and flatten out the bind array as much as possible to reduce ioctl counts