replace expat with etree when parsing unicodes, image and components

[anthrotype]

https://github.com/unified-font-object/ufoLib/issues/80#issuecomment-428527832

for fetching unicodes, image file names and component references, glifLib is still using an expat parser. For consistency we should use ufoLib.etree API for those, like we do for plists and for GLIF. We could of use the elementtree "target" interface (jut like I did for ufoLib.plistlib), which is supported by both cElementTree and lxml: http://effbot.org/elementtree/elementtree-xmlparser.htm#the-target-interface