Add Rao Blackwellized particle filter

[hidmic]

Feature description

Rao Blackwellized particle filters (RBPFs) essentially split state into a non-linear part and a conditionally linear part (conditional on the non-linear part), and then use some closed form estimator like Kalman filter to estimate the latter while still using a particle filter to estimate the former. It's like a KF embedded in a PF. The interesting bit about RBPFs is that they can help cope with high dimensionality. You can focus your PF where it matters, and let some EKF do the rest. AFAIK https://arxiv.org/pdf/1301.3853 introduced it, but I find lectures on the matter like https://users.aalto.fi/~ssarkka/course_k2016/handout6.pdf easier to read.

How each RBPF decomposes state into non-linear and conditionally linear parts will be application dependent, but Beluga can still provide an idiom or pattern to do this

Implementation considerations

KFCore looks like a good candidate for a KF library.