Sampler rework MVP. New issues and refactoring tasks

[nabriis]

This issue tracks comments primarily set in #356 for future updates. This can be split further as appropriate.

• [x] Unify handling of sampler "state" and attributes like "current_point", "current_loglike_eval" "acc" etc. #379
• [ ] Refactoring wrt. proposal-based sampler including a validate_proposal method instead of proposal.setter #397
• [ ] Refactoring "propose", "evaluate" and "accept/reject" in step method
• [ ] Improve "reset" method. #414
• [ ] Allow target=None (and init vars when target is set) #398
• [ ] Handling initial point. Should it be stored as a sample? E.g. for LinearRTO it does not make much sense. For pCN one could draw from prior. #324
• [ ] Consider fixing found bugs in "Sampler" module. And/or refactor to avoid "scale_temp" needed for tuning.
• [ ] Improve progressbar display by including e.g. acc rate.
• [ ] Implement loading of batched samples.
• [ ] pCN tune method
• [ ] Improve "load_checkpoint". Currently, it doesn't touch initial_point and self._samples[0]. Maybe we should set these two as current_point?
• [ ] (add by @amal-ghamdi ) remove reference to demos/demo31_callback.py. in sub sampler classes if not relevant or update the demo to make it relevant
• [ ] For all the sampler classes in which tune is implemented as 'pass', do we need warning or error? to let the user know tuning is not happening?
• [ ] (added by @amal-ghamdi) implement validate_target in all classes
• [ ] (added by @amal-ghamdi) should step and tune methods be hidden?

[amal-ghamdi]

@nabriis, I just noticed that status of this issue is Done. Should this be in backlog? Thanks!

[nabriis]

I did it now. Odd. I created it via the project plan