Fix Qt4 figure editor color setting and getting

[mspacek]

This fixes #2274. It also permits only MPL colors to be entered in the color lineedit boxes. Previously, it seems Qt4 was dealing with any named colours entered into the lineedits, instead of MPL. This probably came out the same most of the time, but was less than ideal. Now, all colors are passed through MPL, which returns an unambiguous hex string for Qt to use. Also, there was some code duplication in terms of checking for valid colors. Also, when editing the color lineedits, the corresponding color buttons now only update when the lineedit loses focus, not on every keypress. Finally, the mass of Qt objects entered directly into the namespace have been cleaned up.

[mspacek]

I just force pushed a shorter, cleaner set of commits.

[mspacek]

Any particular reason why Travis failed? I'd be surprised if there was any test coverage at all for the qt4 editor.

[mdboom]

That's a failure that crops up most of the time, but I haven't been able to reproduce. Benign as far as this PR goes.

[mdboom]

This looks good on first blush. Since this is GUI-related changes, I think I'd like someone else who uses the Qt editor on a regular basis (I don't personally) to look this over and make sure all of their use cases are still met. Maybe we can recruit someone on the matplotlib-users mailing list for this?

[mspacek]

Sure, I'll try sending a message to the user list.

[mspacek]

BTW, ipython was having similar problems with Travis, minrk thought it was due to timeouts:

ipython/ipython#3641
ipython/ipython#3647 (comment)

[mspacek]

@mdboom I posted a message to the user list a week ago, but no one's replied:

http://article.gmane.org/gmane.comp.python.matplotlib.general/33391

I've been using this branch for about 2 weeks now without any problems.

[mdboom]

Ok -- no need to further hold this up.

[mspacek]

Thanks!