Use file extension from content-type header instead of giving up

[iandees]

Currently, when a source file's content-disposition response header has a file extension that disagrees with the content-type response header we give up and use .txt, which causes confusion later in the download/conform pipeline.

Instead of giving up, this switches to use the file extension calculated from the content-type.

This is meant to handle the errors seen in openaddresses/openaddresses#6925

[iandees]

In the linked PR, the content-type header gave us .json and the content-disposition header gave us .geojson. They diagreed, but resolve to essentially the same thing.