feat!: retrieve all markdown pages

[dandhlee]

Recovered all markdown pages in nested directory structures. This was not correctly retrieved as I was only looking in the top directory.

Few things that's happening (happy to break down, but was hard to do so to have all the moving parts in one picture):

• Keeping track of where pages are retrieved. This is vital as we need to know where markdown pages will need to end up in the TOC
• Merging the Markdown and Package's TOC, based on the two sources pulled separately. I could choose to just move all the markdown files at the root directory, but I think a better experience will be to put the markdown pages where they are located.
• Remove unused markdown pages that are generated. These are redundant API pages that Sphinx generated but are not used by DocFX Yaml plugin.

Only the handwritten TOC is updated and works as expected, gapic-combo and gapic-auto doesn't contain any markdown pages in nested directories.

Some markdown pages might come with a mix of handwritten guides and autogenerated API content. I plan on working with the library maintainers to have this updated throughout (<10) for the purpose of this plugin that separates Sphinx HTML and DocFX YAML content.

Breaking Change: Upgrading to using 3.9. All Python client libraries are able to use 3.9, so this is not a problem, but will be a backwards incompatible change.

Fixes #217.

• [x] Tests pass

[dandhlee]

Please take a look again!

[dandhlee]

Please take a look again!