CPF: Conditional Particle Filter
In lolmid/unbiased_intractable_targets: Debiased particle MCMC with couplings

Runs a conditional particle filter, with or without ancestor sampling

1 2	CPF(observations, theta, model, nparticles, ref_trajectory = NULL, with_as = FALSE)

`observations`	a matrix of observations of size datalength x dimension(observation)
`theta`	a parameter to give to the model functions
`model`	a list representing a model, for instance as given by `get_ar`.
`nparticles`	number of particles
`ref_trajectory`	a reference trajectory, of size dimension(process) x datalength; if missing, runs a standard particle filter.
`with_as`	whether ancestor sampling should be used (TRUE/FALSE)