FeatureRequest: Migrate to a faster xml parser

[groceryheist]

Others have reported improved performance when using expat to parse Wikimedia dumps. We are currently using ElementTree which provides a good balance between usability and speed.

There is probably potential to speed up this library by switching to a faster xml parser. Candidates include:

• lxml
• cElementTree
• expat

Migrating to lxml or cElementTree might be relatively easy because they have similar APIs to ElementTree.

[halfak]

I tested cElementTree a while back and found that I generally got similar or worse performance in python3. I'm not sure why. That was over 2 years ago, so it might be worth testing again.