Switch to WaniKani Image Mnemonics project for subject artwork

[aaron-meyers]

As first proposed in comments on #653, let's switch to using the WaniKani Image Mnemonics community project for subject artwork. This project aims to create artwork for all subjects in WaniKani (radicals, kanji, vocabulary) and has made a lot of progress using recent AI image models like DALL-E 3.

This should be a fairly straightforward change to replace the hard-coded AmandaBear artwork URLs with the community project CDN images as implemented by this userscript: https://greasyfork.org/en/scripts/448713-wanikani-ai-mnemonic-images

More info on the community project here: https://community.wanikani.com/t/new-volunteer-project-were-using-ai-to-create-mnemonic-images-for-every-radical-kanji-vocabulary-come-join-us/58234

Part of #689 support for WaniKani Image Mnemonics subject artwork.