We previously used Python-Markdown's default behavior of guessing

the language based on the contents of the code. This was often wrong, or

didn't result in anything meaningful. Fortunately, that almost never

came up in usage (since we were using marked.js and

google-code-highlight for the rendering), but when we move to

server-side-only rendering of Markdown content, it'll become more

important.
Now, syntax highlighting is only enabled if using fenced code blocks

with a language specified.
The manual has been updated for these changes.

Tested locally with a bunch of review requests. All old code blocks

without fences and a language identifier appeared as plain text, whereas

those with a language identifier highlighted properly.
Built the docs and read through them.