Refactor of AST traversal method

[abelsromero]

Thank you for opening a pull request and contributing to asciidoctor-maven-plugin!

What kind of change does this PR introduce? (check at least one)

• Bugfix
• Feature
• Documentation
• Refactor
• Build improvement
• Other (please describe)

What is the goal of this pull request?

Solve limitations in the current method used to traverse the AST.
Currently, we explored all nodes in a loop in NodeSinker that already showed issues with nested elements that need to generate content "around" nodes, like lists. That forced to have specific loops for such elements.
The new method does not look externally (outside of NodeProcessors) but instead injects NodeSinker in the NodeProcessors so each one decides if further traversal needs to happen. This has simplified the code notably and removes the need for specific dependency injection.
Now all NodeProcessors are instantiated and stored in NodeSinker without special cases.

Are there any alternative ways to implement this?

Probably, so far this works 😄

Are there any implications of this pull request? Anything a user must know?

There are some changes in how HTML content is generated:

• Empty document no longer generated empty <h1></h1>
• Empty literal no longer generated empty <pre></pre>
• Sections are now wrapped around <div> elements for better organization.
• Fixed Listing blocks outputting content instead of source

Is it related to an existing issue?

• Yes
• No

But it's related to this PR #938

Finally, please add a corresponding entry to CHANGELOG.adoc