Keep emphasis nested directly in emphasis when re-rendering Markdown - #312

MarkdownRenderer writes every emphasis with *, so when emphasis is the only content of another emphasis, the delimiters of both join into one run and the nesting changes. The common bold italic **_foo_** is one case:

>>> md = marko.Markdown(renderer=MarkdownRenderer)
>>> md("**_foo_**\n")
'***foo***\n'

<!-- before -->
<p><strong><em>foo</em></strong></p>
<!-- after re-rendering -->
<p><em><strong>foo</strong></em></p>

The same happens with __*foo*__, and *_foo_* (spec example 460) and _*foo*_ (462) come back as **foo**, which is strong emphasis.

When the only child of an emphasis or strong emphasis is an emphasis, that inner emphasis is now written with _ (or * if the outer one uses _), so **_foo_** stays **_foo_**. Its _ always has the outer * next to it, so the rules for _ inside words don't apply. Other nesting is unchanged: ***foo***, emphasis around strong emphasis, already gives the same HTML, and so does foo***bar***baz.

Added test_markdown_renderer_nested_emphasis; 5 of its 7 cases fail on master, and the last two check that ***foo*** and foo***bar***baz are unchanged. Across the CommonMark spec examples, the number whose re-rendered Markdown gives different HTML goes from 17 to 15. pytest tests, black, isort and flake8 pass, and mypy reports the same 5 errors as on master.

Used Claude (Opus) to help find and patch this; I reviewed and tested the change