Replace string manipulation with AST logic in parseArgReferenceSection

[guineveresaenger]

Uses Goldmark to extract information for docs descriptions from Markdown nodes, specifically:

• ListItem
• top-level Pragraph
• Section Headings

This PR preserves the logic of tracking a docs path via number of subfields for later lookup, and assigning it a description.
It also attempts to keep all of the docs parsing details of before.

This code is adjusting the azurerm-sql-firewall-rule test because it doesn't yet preserve the **NOTE** inside of an argument description. (Arguably, this NOTE should not actually be part of an argument field - it should exist on its own as part of the Section.)

Additionally, H4 headers are now correctly nested under H3 headers. The example for the TestArgumentRegex/Parses_H3_and_H4_headers_and_their_bullets_as_nested_properties test
is taken from here and it should in fact be nested in the way the new changes reflect. Ultimately for docsgen it doesn't seem to make a difference, but it is nice to see that structures are preserved.

Currently there's a few subheading patterns that aren't carried over to the new parser quite correctly.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 57.00%. Comparing base (5548bd6) to head (45847da).
Report is 14 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2318      +/-   ##
==========================================
+ Coverage   56.85%   57.00%   +0.15%     
==========================================
  Files         364      365       +1     
  Lines       49910    49943      +33     
==========================================
+ Hits        28374    28470      +96     
+ Misses      19990    19919      -71     
- Partials     1546     1554       +8     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.