Convergence diagnostics

[jeffjennings]

Add plotting for convergence diagnostics, other various minor changes

Fix #45 Fix #46 Fix #47 Fix #49 Fix #50

@jeffjennings: the "Fix #N" statements I added yesterday in this description were not reminders to be removed.

A "Fix #N" in a PR description or in a commit message will close the #N issue when the PR is merged. It is a bad practice to close the issues manually, and it is actually better not to unless they become irrelevant, and won't imply any change to the repo.
Using the "Fix #N" statements ensures that the issues are closed really when they are closed (i.e. when the PR is merged). Until this PR is merged the issues are not closed. If we were not to merge this PR now, you would have closed issues that are not actually closed, with the obvious confusion.
Secondly, by using "Fix #N" statements Github automatically adds a reference with a link to the PR or the commit where the issue was closed, in the issue page, which helps a lot debugging later on. While now you added manually "see file..." without the PR reference number, which makes it very hard to track the commit down

[rbooth200]

This PR is ready to go from my point of view.

[rbooth200]

Marco tells me @jeffjennings is still working on this, I'll wait to close it.

[rbooth200]

I've fixed some issues that pylint was giving, but the diagnostic plots aren't working yet, apparently.

[rbooth200]

Quick Q: why did you remove the pylint directives?

[jeffjennings]

My mistake, I thought it had automatically added those and that they could be removed. I've just pushed with them added back in.

[jeffjennings]

Everything works for me on this branch now, including the diagnostic plot.

[rbooth200]

I'd also suggest replacing start_iter and stop_iter with a single num_iter parameter and just print the last so many iterations. Isn't that what will be wanted anyway? I think there is too much chance of error with the current parameters.

Obviously if you ask for 1000 and it converged in 20 then we just show all of them.

[rbooth200]

The pylint directives are to stop it reporting false positives

[jeffjennings]

Hey @mtazzari, @rbooth200 and I have both been contributing to this branch that closes the outstanding issues in the Code freeze milestone. Can you review and merge into master? Thanks!

[mtazzari]

Ok, starting the review

[rbooth200]

I've just merged in some changes needed to get tests of the full pipeline working

[mtazzari]

Some small changes requested.
Ok to merge after that.