[Suggestion] Use a Commonmark parser for the rendering of readmes

[stof]

Currently, the rendering of readmes has 2 different paths:

• using the Github rendered readme for github repos, which ensures that we get the same parsing than on Github (as Github is doing it)
• rendering the markdown with cebe/markdown for other packages (as we don't have pre-rendered readme for them)

This suggestion is about replacing cebe/markdown for the second case.

Nowadays, most non-github repositories on packagist are probably Gitlab ones. And Gitlab's markdown is based on CommonMark (with a few extensions of their similar to the Github ones, see https://docs.gitlab.com/ee/user/markdown.html). And Bitbucket also uses CommonMark (https://confluence.atlassian.com/bitbucketserver/markdown-syntax-guide-776639995.html).
So using a CommonMark parser would probably be better for compatibility of the rendering. But cebe/markdown is not such a parser.

[stof]

Another idea: make another special case for the rendering for Gitlab, as they now have an endpoint to render markdown: https://gitlab.com/gitlab-org/gitlab-ce/issues/24124#note_131775879

[stof]

Note that the go-to commonmark parser for PHP is https://commonmark.thephpleague.com/

[colinodell]

FYI league/commonmark now includes an extension which provides full GFM support: https://commonmark.thephpleague.com/1.3/extensions/github-flavored-markdown/

This 100% meets the GFM spec, meaning that what we render should be nearly identical. (The only difference is that Github's Markdown API has some extra post-processing in their HTML pipeline, to do things like render emoji)