Double image proxy confuses bypass logic

[Sjmarf]

E.g. https://lemmy.ml/api/v3/image_proxy?url=https%3A%2F%2Fsopuli.xyz%2Fapi%2Fv3%2Fimage_proxy%3Furl%3Dhttps%253A%252F%252Ftenor.com%252FVdlE.gif

I'm guessing that we just need to modify whatever piece of code is extracting the original URL from the proxy URL to consider the possibility of there being more than one proxy.

[EricBAndrews]

MediaLoader is supposed to handle this as currently written by recursively stripping the last proxy bypass parameter off the url and re-calling load with that new url. It looks like computeProxyBypass isn't behaving as expected.