Add support for images stored in memory (Images downloaded from a url)

[nleroy917]

I understand the concern from earlier where Colorgram was responsible for downloading the image, this time I made it so that one can downloaded the image with requests and pass the data in memory to colorgram which detects data type and opens the image accordingly.

[obskyr]

Hmm, I don't dislike this quality-of-life change – not having to import PIL is nice – but…

• The code has a bunch of references to "URL" now – this change has nothing to do with the internet, so I'd like different names for those.
• test_api.py shouldn't depend on requests. Instead, try opening one of the local images, reading that as binary data, and passing that!
• I'd also like a test in test_api.py for passing a file path to extract, since that's something passing the argument directly to Image makes possible.
• Do please squash it to a single commit, or at least ones that don't include the old request-including commits!

[nleroy917]

Sure - this will require another pull request