Support (and ignore) commonmark header attribute lists in SUMMARY.md

Problem

mdbook supports CommonMark header attribute lists, and handles them (particularily the #id) attributes appropriately. However it does not handle them in one particular place - SUMMARY.md.

This means that they show up in the table of contents sidebar when they are used (for example, to adjust an identifier computed for parts by 3rd party renderers).

Proposed Solution

The {attributes} block in a h1 header in SUMMARY.md should be parsed by mdbook's html renderer and discarded. This allows other backends to make use of this block while not being visually intrusive in the default renderer.

As an alternative, though one that would likely require an API breaking change, the attributes could be provided separately to the string. This would make it easier for both backends that make use of it (not requiring the backend to explicitly parse it) and ignore it (being able to discard it silently without effort. The pulldown_cmark crate provides this information, so this would likely just require enabling that feature in it.

Notes

In https://github.com/chorman0773/mdbook-fiction-tools, the "ID" of each part is used to compute the output file for the epub output (as well as others using the same system in that repository) when the user of the tool requests that output be split by part (where multiple different works may be amalgamated into a single repository and rendered to html using the same mdbook book, but rendered to downloadable formats elsewise). It also uses it to determine keys in the file-ids config table, which allows users to set explicit identifiers per-part.

Trace how h1 headers in SUMMARY.md reach the HTML renderer, then inspect the pulldown_cmark attribute support mentioned in the issue. Done means attribute lists are omitted from the rendered sidebar while remaining available to other backends; add or update coverage for SUMMARY.md parsing.