Fix exposure logging + cleanup

[eytan]

Logging conditional on whether there is a matching variable can cause downstream (errors in randomization)[https://github.com/HubSpot/react-experiments/issues/1]. This diff fixes that by not making logging dependent on having a matching variant. This diff also includes a number of changes that increase specificity, including:

• Not making logging depend on whether there is a matching variant.
• Renaming variables to better between variant names, exposures, and selected components
• Renaming Experiment to VariationExperiment, since this component reflects a more limited type of experiment (a simple A/B test, or "variation experiment") that people may want to run. This also helps distinguish itself from PlanOut.js' experiment class
• Improvements to index.html: have exposure logging print to console for easier debugging / understanding; using lowercase variable name to refer to a class instance
• Added more specific development setup instructions -- it seemed that one has to do an npm install for babel and related libraries for things to run
• The web pack build was pointing to ../dist, so that running sh build.sh would build to a directory above the repo. This makes it point to ./dist.

Todo:

• Nuke utils.js
• Possibly get rid of experimentEnrollment. It seems to add a lot of clutter and is mostly there for doing conditional logging.
• Rename experimentClass prop to experiment, since it expects an instance of an experiment class
• In the future, we should include a legit propType for the VariationExperiment class that corresponds to a PlanOut.Experiment object (I tried doing this before but it didn't work)
• Mostly neglected the namespace class because it sounds like something that Guy is going to nuke.

This reverts commit c85a224531d4e21d0220b75b8dc0b9934071c6f1.

[eytan]

@rawls238, does the list of TODOs seem reasonable to you? I was expecting that this diff might be a little contentious... After more further consideration, my plan would be to get rid of the experimentEnrollment object, as well as the isEnrolled property.

[rawls238]

I think that the experimentEnrollment class is still needed as long as the Namespace class is still in use. The process of writing up a few changes to the overall interface in #1 may make it necessary to repurpose this class for different uses.

I think that the changes to logging make sense. We originally added the conditional logging because we had a few experiments where each variation was in a separate component (though, if everything was working correctly in the code, an exposure event would log for each variation since they were all in the same area on the screen on the same page). The idea was that having exposure events only log when the variation was actually rendered would theoretically allow us to be able to detect when something is wrong with one of the variations, since we would not see the right # of exposure logs for that variant.

This method opens up a huge can of worms (for instance, the problems that you mentioned) that I'm not sure how to prevent so just doing automatic exposure logging makes sense as one piece of the solution. I think a decent compromise may be to log an exposure event as done in this PR, but also log an experiment event (using logEvent) when the variation component is actually rendered to get the best of both worlds. I think it's important for this library to work when experiment variations may need to be defined across different components so we should aim to ensure that we can allow consumers of the library to do so regardless of how their existing code is structured and without making it easy to shoot yourself in the foot. (Let me know if you disagree with anything here / find any flaws in it.)

On that note, we should nuke utils.js. Also agree with renaming the experimentClass, but not sure we should add propTypes for the experiment prop since I still want to allow consumers of the library to be able to use their own experiment Classes.

[eytan]

OK @rawls238, here are my changes. Probably won't have much time this week to work on it much more, so you could either merge and fix up later, or take whatever parts you'd like and I can abandon the PR.

[rawls238]

thanks for doing this @eytan, I've added some of these suggestions to https://github.com/HubSpot/react-experiments/pull/6 but for the most part since the library has effectively been rewritten most of these changes are now irrelevant so closing this