be open-minded about byte and dict headers

[wumpus]

This is as discussed in #16

I kind of punted on converting bytes/str/whatever for Python 2 because I don't really understand it. It still does dict -> list for python2.

If the user happens to have multidict installed (part of aiohttp), then the multidict tests will run. This is optional because warcio strives to avoid external dependencies. If there's a more-correct way to get that into the CI, I'd love to know it.

[coveralls]

Coverage decreased (-0.1%) to 99.903% when pulling 5485134 on cocrawler:header-frobbing2 into 173aa90 on webrecorder:develop.

[wumpus]

I don't understand the coveralls fail. The marked line is executed by both python2 and python3.

[wumpus]

shit. missing a file.

[coveralls]

Coverage remained the same at 100.0% when pulling f8ccd14 on cocrawler:header-frobbing2 into 173aa90 on webrecorder:develop.

[coveralls]

Coverage remained the same at 100.0% when pulling ff079b2 on cocrawler:header-frobbing2 into 173aa90 on webrecorder:develop.

[wumpus]

Oy. Hit the "squash merge" button :-)

[coveralls]

Coverage remained the same at 100.0% when pulling 72ef825 on cocrawler:header-frobbing2 into 173aa90 on webrecorder:develop.

[ikreymer]

Thanks for this, and sorry for the delay.

I wonder if its worth optimizing for headers already being in str format:

if h and type(h[0][0]) == str and type(h[0][1]) == str:
   return h

I suppose its unlikely that there'd be a mix of str and byte headers, but not sure if this would really make any difference