BTW: There seems to be a failure with 5 kml tests that already failed in the current master. I had hoped this was due to my local setup, but it seems to fail on Travis as well. I am not really sure why this happens and I think this should not be related to this PR.

I think the CI failures are caused by 9c5c3cc. In fact, #123 shows that the author is aware of the fact that the test data should be updated. Unfortunately, this has not happened.

I think it makes sense to rename this from origin to something like gpxType. That way, it is more intuitive for togpx to support the same tag so that a roundtrip togeojson-togpx retains as much of the original as possible. For the togpx use-case, the name origin would be a bit weird.

The togpx part of this change would then be an alternative to tyrasd/togpx#11.

Sure, we could change the name. So far this PR did not receive much attention, so I am not sure if it's worth the effort?