batch_filter modifies kf values

[rlabbe]

I'm not sure if I like this or not. The function signature implies that x,P are computed and returned in a list, but the kf states are also updated. Maybe that's okay, maybe not, but it is not documented!

It strikes me that it should not modify the kf. Perhaps it shouldn't be a kf method at all, but a stand-alone function that takes the kf as an argument just to get the matrices.