Storing statistics on the particle weights

[trappmartin]

As discussed in TuringLang/Turing.jl#426, it would be good if Turing would be able to store statistics on the particle weights for all particle-filtering based methods. These statistics could be stored in the ParticleContainer and exposed to the user through new fields in the Chain type.

[fipelle]

Any news on this? The output of sample(...) seems to be the final snapshot of the particles and weights.