If the drone is moving in 6DoF, setting Reg/Force3DoF to true would probably not work.
You could however use
robot_localization package to fuse the parameters you want with sensors you want. You may want to only fuse twist from VO into that filter (in case there is a drift in roll or pitch).
Note however the PX4 EKF is kinda doing the same thing than robot_localization EKF filter, though maybe not as configurable. Anyway, I think any "vision pose" you send to PX4 EKF (even if coming from another EKF) will be fused again by PX4 EKF with the onboard IMU.
Note that I have never tried to fuse another source of odom, so I cannot really more than that.