Online thinning of samples

[amal-ghamdi]

Is your feature request related to a problem? Please describe.

In image-related problems, we may need to run long chains of millions of steps. Storing all intermediate samples is computationally expensive and often unnecessary, so implementing online thinning would be beneficial.

Describe the solution you'd like

Ns=100
Nt=10 # Nt is thinning interval/frequency
sampler.sample(Ns, Nt)

Related to #691