Improved JavaScript extraction

[akx]

This PR improves the JavaScript message extraction system, adding support for

• dotted names as keywords (com.corporate.i18n.gettext) (#148)
• tokenization of JSX/E4X <tag>s (to avoid misparses) (#280)
• ES6 template strings in two different flavors (#329)

It's a refactor and superset of the previous JavaScript PRs, so:

Closes #298
Closes #310

[codecov-io]

Current coverage is 88.97%

Merging #332 into master will decrease coverage by -0.59% as of 2117212

@@            master    #332   diff @@
======================================
  Files           23      23       
  Stmts         3785    3800    +15
  Branches         0       0       
  Methods          0       0       
======================================
- Hit           3390    3381     -9
  Partial          0       0       
- Missed         395     419    +24

Review entire Coverage Diff as of 2117212

Powered by Codecov. Updated on successful CI builds.

[zackargyle]

This looks fantastic. Thanks for adding further support!

[catmeme]

Any progress on this? I am in need of dotted names as keywords. Thanks.

[akx]

@catmeme If you feel like it, could you give the PR a quick code review pass?

[RauliL]

LGTM.

[akx]

Merging. (@RauliL is a colleague who I recruited to do CR.)

[zackargyle]

@akx, any chance you could update the pip package? We're ready to try out the new JSX extraction!

[lyapun]

@akx Yeah, we are waiting for this too :)

[sassanh]

@akx thanks for this great package. Unfortunately seems like the jsx support is not complete, it doesn't work on something like this:

<input ref='login-email' placeholder={gettext('username')}/>

is it the same for you or am I missing a config?

[sassanh]

@akx nevermind, I downloaded latest dev version (2.4.0-dev) and it's fixed.
It'd be great if you'd add make install in setup.py so that when someone installs a git commit using pip it generates locale-data.