SMC² State Particle Time Indexing

[charlesknipp]

In the newly implemented SMC² algorithm, the indexing of state particles is rather odd. It will propagate normally on the first few steps, but after awhile with sufficient parameter particles, the filter will have gone through far more iterations than the parameter particle time index suggests.

Essentially problems arise because smc².filters.state.t[] is far greater than the "supposedly" equivalent smc².parameters.t[]. This could be a multitude of reasons, but my guess is the indexing of Reference objects is confused when changing an immutable object.

This is a problem I have had to deal with before, but have never found an apt solution which does not involve a complete redesign of the particle filter object.