Move visualRegression subtask from npm scripts to grunt task.

[cjcenizal]

This PR supersedes #7461.

Changes

• Log dependency path info in licenses grunt task. Now you'll see a path of the parent dependencies when a dependency fails to licenses task, so you can identify the offending root dependency.
• ES2015ify the licenses task.
• Add an override for abbrev@1.0.7 because the licenses task was failing for that dependency, even though it has the MIT license.
• Remove test:visualRegression from test npm script. Add it to end of test grunt task.

[ycombinator]

I'm not familiar with how we decide whether its okay to add a package to this list. Perhaps @spalger or @epixa might know?

[spalger]

It's based on the package having a license in an undiscoverable format. Like, if it's mentioned in the README somewhere, or in a github issue, the license discovery tool won't be able to detect it automatically so we hard code it here

[cjcenizal]

I'm actually not even sure if this is the right way to handle this... I tried to figure out the root cause of the problem (since this is only happening to me apparently), but gave up and figured it's better to just sidestep it and move forward. 😠 Should I go back and work this bug over instead?

[cjcenizal]

I also just realized the license is ISC, not MIT.

[spalger]

Cool, as long as you add the correct license to the overrides we can avoid calling Septa Unella

[ycombinator]

The dependency path tracing is very cool! Correct me if I'm wrong but it seems unrelated to the change of moving visualRegression from the npm script to the grunt task, as a subtask. If so, consider making a separate PR for the dependency path tracing changes (or updating this PR to be about those changes and make a separate PR for the visualRegression changes).

[jbudz]

@cjcenizal I looked into this a bit more and updating our npm dependency does seem to fix the license check issue. Can we open a separate PR to update npm to 2.15.1(and close #7184) and keep this PR for moving npm scripts?