Add BaseExperiment._finalize method

[chriseclectic]

Summary

This adds a _finalize method to BaseExperiment called in run before the actual experiment is run which gives users a hook for setting any custom experiment or option values based on the final configuration of the experiment before execution.

Details and comments

The main goal for this is to allow developers/users to be able to set automatically calculated experiment or analysis option values that might depend on a combination of other option values when writing an experiment subclass. For example it could be used to configure a specific data processor in analysis based on the backend and run option values (@eggerdj), or to set the initial guess value for CR hamiltonian in #695 (@nkanazawa1989)

[eggerdj]

This looks like a nice addition. It might also help us implement restless measurements which require a modified data processor based on the run options. See e.g. #689. @chriseclectic aside from the code, is there a place that describes the workflow of the base experiment (e.g. a flow chart showing the conceptual steps) that also contains all the possible hooks?

One question: how does this work with parallel experiments?

[nkanazawa1989]

This looks good to me. To add extra metadata to transpiled circuit as done in the standard RB experiment, a subclass still need to override _transpiled_circuits method rather than passing circuits to this method?

(EDIT)
Probably this can be called before _transpiled_circuits since some experiment may want to override transpile options?

[chriseclectic]

@eggerdj Good point for composite experiments. I believe that should work correctly if composite experiment implements a finalize method that recursively calls finalize for each sub-experiment and that composite experiment set_run_options is updated to recursively set run options in each sub-experiment.