[WIP] Add SMC init support/progress bar with kernel ops tracing

[rrkarim]

PR adds support for Sequential Monte Carlo (SMC) method API. SMC is used to tune the step_size of NUTS sampler.

SMC support:

• [x] Support SMC initialization for mcmc sampler pm.sample(model(), initialize_smc=True/False)
• [x] Provide notebook of Jun's work on SMC (partly done)
• [ ] Add comprehensive tests because of the modified logic of the graph executor (this is important)
• [ ] Add similar notebooks as in pymc3
• [ ] Add support for num_chains (I've runned to some issues but will solve this soon too, this may cause high variance in samples)
• [ ] Add support for other kernels (This one is maybe not possible (or hard) with our design of bijectors, i.e. bijectors in pymc4 are determined by executing the graph)

Progress-bar support

• [x] Add the support for progress-bar for better debugging pm.sample(model(), progressbar=True/False)
• [x] Add notebook with the performance check (not the ideal way of measuring the performance though
• [ ] Add tests

Also PR adds support for mixture, but there is a conflict with #284 (added to test on mixture likelihood model for SMC)

[review-notebook-app[bot]]

Check out this pull request on 

Review Jupyter notebook visual diffs & provide feedback on notebooks.

---

Powered by ReviewNB

[junpenglao]

Seems there are some conflict and a lot of files shows up - could you upstream to master branch first?

[rrkarim]

Seems there are some conflict and a lot of files shows up - could you upstream to master branch first?

yeap, solving this now.