Clarify webdriver / chromedriver and screenshot debugging

Following up from work I've been doing with mass.gov:

chromedriver is supposed to speak the webdriver protocol. selenium is also the webdriver protocol. So why do we need separate base classes? Does chromedriver speak webdriver, but also extend it with additional non-webdriver APIs like the call to http://chromedriver:9515/json/version?

After digging into it, I determined that the "WebDriver" base class does not really speak the webdriver protocol. Eventually, I think we should break compatibility and rename all of these classes. I also wonder how many teams are using the chrome debugger protocol, as I haven't seen any ARM container images yet that offer Chrome and a VNC server, whereas Selenium does.

Regardless, this updates documentation only, as well as adds a note about how screenshots can make tests pass unexpectedly.