Do not expose arbitrary HTML ids from Markdown headings

A markdown cell with the content "# xyz" will create HTML of the form <h1 id="xyz"> ....</h1> if there is more than one word, the id will use hyphens in place of spaces.

This can lead to very surprising problems when a use chooses a heading word that is used as an id by another javascript/HTML component for their own purposes.

This was the case with Bokeh, see https://github.com/bokeh/bokeh/issues/10002#issuecomment-654810980

I think it is a dangerous practice in general to use this approach which can result in arbitrary ids without any jupyter specific prefix. Especially in notebooks about some other library (like Bokeh) it can easily happen that such a library specific id is mentioned in a heading. Of course all libraries should have some convention to prevent clashes of ids (and class names).

I think this is related to, but different from #1899

No file or test is named. Start by locating the Markdown-to-HTML heading renderer and reproduce the # xyz case, then inspect the related #1899 discussion. Done means heading IDs no longer create arbitrary collisions with notebook or library HTML/JavaScript IDs, with regression coverage for the reported Bokeh-style case.