Adjust logic for selecting top-level sections

[ghyman-oreilly]

Hey team, this PR follows on #71, which I posted after trying to convert a bunch of notebook files where the section/chapter wrapper didn't match the pattern used in get_top_level_sections.

With this adjustment, in instances where a wrapper isn't found, the function will try to grab a list of top-level sections (currently we list all sections, I think). If there's more than one top-level section, the existing warning is logged; otherwise, we can proceed with processing the single top-level section.

Hopefully, most of the the time the notebooks we're working with will follow the expected form, but I've seen this week two patterns (albeit in the same project) that were well formed overall but not as expected:

• article wrapper is present but has no role attribute
• wrapper with role of main has tag type of main rather than article

PR includes a test case with data matching the former of the two above patterns. Let me know if you have feedback or want me to make any changes. Thanks!

[delfanbaum]

Nice work! Looks good, thanks for including what edge cases the test is looking at in the docstring (since it's a pain to always go digging through the HTML).