Hello.

I've got a seemingly working version of completion of HTML class names
and IDs in CSS mode buffers.

The idea is that CSS mode asks other open buffers if they are able to
produce completion candidates for HTML classes and IDs by checking
whether the functions `html-class-extractor-function' or
`html-id-extractor-function' are bound. I've included implementations
for these in HTML mode using libxml. My hope is that other modes, such
as Web mode [1], will be be able to implement their own extractor
functions.

What do you think?

-- Simen


[1] http://web-mode.org/