Infer and allow BBcode lists [ul] in class reference generation

[Mickeon]

Supersedes #67037

I've had these two screenshots on my desktop for the past 2 weeks and I forgot to create this PR:

This PR allows the [ul] to be used in custom script's class reference generation. However, unlike #67037, it also infers a list based on whether or not a line starts with -.

Why like this?

• It does not require changing all existing docs that already do this for a list
• It is familiar syntax (this is how I'm writing this list on GitHub right now).
• The online class reference generated by make_rst.py already does this, whether it's intentional or not.
  

(Only reason this is marked for 4.3 is because the prior PR was, too)

[Calinou]

Tested locally, it works as expected.

Note that nested lists are not supported by this code though. The following in the XML:

- List
  - Nested
  - Item
    - More nested
  - Less nested
- Even less nested

Results in:

I don't recall seeing nested lists in the XML class reference, but it's worth keeping in mind for custom class reference from scripts and should be documented once we make a PR for godot-docs.

[Mickeon]

The moment nested lists will be supported is when the class reference will adopt the same syntax as GitHub's markdown which would be... wow.
Indeed there's no "nested" lists and I believe the docgen would catch and error if spaces are used to pad the beginning of a line (unless in a codeblock).

Out of guilt, I must note that there is a situation where the current code would break: since [codeblock] is not checked on the [ul] conversion pass, if a codeblock line were to start with - , it may break. As far as I know the class reference never does lists like this, but it is a possibility.

[dalexeev]

• It does not require changing all existing docs that already do this for a list

I think it's better to update the docs and not add the inferring. Or this should be done not by a simple replacement, but take into account the context of [code], [codeblock], [ul] and [ol] (not implemented yet) in the EditorHelp BBCode parser.

## [codeblock]
## - A
## - B
## - C
## [/codeblock]
func _run() -> void:
    pass

• The online class reference generated by make_rst.py already does this, whether it's intentional or not.

This is due to incomplete RST markup escaping.

[Mickeon]

I'd rather have it account the context better, then. But in all honestly the generation code is quite "spaghetti" as is and I wish it weren't so this addition could've been integrated more naturally.

[dalexeev]

I'd rather have it account the context better, then.

I'm not sure if multiple ways to do the same is a good thing. RichTextLabel already supports [ul] and nested lists, unlike the new syntax. Yes, EditorHelp provides some shorthands like [param name] instead of [code]name[/code], but here you are mixing BBCode with Markdown.

[dalexeev]

• Needs work. Personally I would prefer Add support for [ul] tags in documentation #67037.

[Mickeon]

How comfortable would you be if this came along with some form of support for GitHub Flavoured Markdown-styled writing for the class ref?

I myself right now am heavily in support of the PR as is, assuming it is done in the least "hacky" way. I want existing documentation to be able to remain as is, and I want documentation to be easy to backport.