Ignore base64 encoded data URLs

[ags]

This removes base64 encoded src values from img tags, which are not widely supported by markdown renderers, are are causing parser performance issues and crashes.

[TJC]

Would it be worth emitting a deprecation warning when base64-encoded data URLs are encountered?

[ags]

Not really a deprecation if they're removed entirely already; we could have done that in an earlier 1.x release, but since this is a 2.x breaking change I don't think it's a problem

[mattptc]

I ran

SELECT content_id
FROM drafts
WHERE document_json->>'body' LIKE '%base64%';

and didn't find any use of this in drafts. Are we seeing used anywhere?

[ags]

and didn't find any use of this in drafts. Are we seeing used anywhere?

This gem is used in the draft editor for converting Word documents to Markdown, prior to the content being persisted. We've seen it happen a few times with one of the FR freelance editors copying from Word, but the conversion failed because of these image URLs and locked up the request worker while trying to parse it (we had a brief editorial-only incident the other day because they tried repeatedly and locked up many workers). I won't share it publicly, but there'll be a PR for the CMS to upgrade this gem where I'll include the Word doc from FR that triggered the problem