* Replace dynamic regex with string find operation
* Add cache of where each line starts so we don't have quadratic behavior identifying line numbers when importing large chunks of html

The origins of these are three-fold:

* Merging in stubs from https://github.com/python/typeshed/tree/main/stubs/Markdown using "merge-pyi"
   - Note: we can consider these annotations to be the important ones because it's what people have been adding according to their own need
* Double-checking around places where stubs were already added from the above, particularly conflicts with annotations that got added in this repository already
   + Taking the opportunity to declare a generic "Registry of T" class
* Running mypy and eliminating the most glaring errors it reported

Every edit to the changelog must be validated so that an invalid changelog doesn't block a release.