GeezerVison

Adds a MarkdownRenderer that renders a Node tree to CommonMark Markdown. Supports all core nodes as well as "gfm-strikethrough", "gfm-tables" and "ins" extensions. This is very useful for applications that want to produce Markdown, e.g.:

• Parse some input Markdown, modify the nodes (AST), then render back to Markdown
• Create a Node tree via code and then render that to Markdown

The goal is that Node tree -> render to Markdown -> parse to Node tree results in an equivalent Node tree. That is verified with all the examples from the spec and some manual test cases.

However it does not mean that a Markdown -> parse to Node tree -> render to Markdown round-trip produces the same Markdown as the original. The parse step does not preserve enough detail for that yet, e.g. things like insignificant whitespace or what characters were escaped.

See #12.

In the future, we can make changes that preserve more details when parsing and make use of that to bring the rendered Markdown closer to the original. But the Markdown renderer can't rely on those extra details to work correctly, as that would break the use case where there is no parsing (when creating Node tree via code).