list by length before processing the element tree

This ensures that multi-word abbreviations are
implemented even if an abbreviation exists for
one of those component words.

To test this,
added `test_abbr_superset_vs_subset()` to
`tests.test_syntax.extensions.test_abbr`.

handling for empty abbreviations.

- Configuration option `use_last_abbr` allows using
the first instance of a term rather than the last.
This is useful for sites that auto-append a glossary
to pages.
- If an abbreviation does not have a definition it
will skip the term instead of writing abbr tags with
an empty title.

The glossary file also uses the Markdown
abbreviation syntax (`AbbrBlockprocessor` is
used to process the file) and keeps the glossary
definitions separate from the page definitions,
allowing the glossary to be applied to every page
while only being processed once.

to accept a dictionary instead of a file.

Changed `load_glossary` to merge in a dictionary,
overwriting any existing records.

Added a `reset_glossary` method to clear the glossary.

Removed the `use_last_abbr` option, since `glossary`
can be used to resolve the use case `use_last_abbr`
was written for.

Completed removal of artifacts from removing
`use_last_abbr` and updated documentation.

Also added testing to avoid having an abbreviation
where `{"": "Definition"}`.