A while back we added an escape hatch for mapping certain file

extensions to mimetypes for when mimeparse messes up (specfically with

text/x-rst and text/x-markdown). While I was looking at this code, I

realized that we go through all the trouble of using mimeparse only to

throw away the result. The mimeparse code isn't doing a whole lot, but

it's always better to avoid doing work if we don't need to. This change

swaps those so if we're going to match based on the file extension, we

do that first, and only use mimeparse otherwise.

Ran unit tests.