For ease of future management, keep each VM in a folder of its own (the entire thing), unless you have a really good performance reason to not do that.
Personally, in your case, I would not split OS and data disks of VMs between partitions, but entire VMs per partition. This way if you need to migrate a VM for performance reasons, you can.
Your big impact will be SQL Servers. And their data disks. So why place them in a configuration that gives the maximum cross VM disk IO impact? Putting all the data VHDs on one RAID partition, simply makes them all impact each other.
Instead of SQL VMs A and B on RAID LUN 1 and SQL VMs on C and D on RAID LUN 2.
In this second model only A and B impact each other and only C and D impact each other. With all the data disks on one RAID LUN, all SQL VMs impact all other SQL VMs and you actually end up in a situation of running out of IOPs much, much faster.
Plan for spreading entire VMs based on IOPs, don't get all fancy with spread around configuration files and OS and data disks. That is not designing for ease of management or making future changes.