Preserve users' intentions during concurrent rich-text editing (refer to Peritext)

[chacha912]

Description:

The Peritext article analyzes nine examples of rich-text editor behavior in practice, forming a model to preserve user intent during rich-text editing. I've tested these examples using Yorkie(in peritext-example-test branch), and I'd like to share three cases that encountered issues:

1. [ex2] Document does not converge during concurrent formatting and insertion

• Description of test case:
  

• While the second result better preserves the user's intent, both of these merge results can be considered correct, as long as both users converge to seeing the same document. However, in Yorkie test, the final documents are not identical.

  https://github.com/yorkie-team/yorkie-js-sdk/blob/2f067ba23994c10ce8791109e0bd915f2e05e62d/test/integration/text_test.ts#L443-L453

2. [ex7] Case where a single character can have multiple mark values, such as comments

• Description of test case:
  

• Yorkie doesn't provide a method to retain both comments.
  Introducing mark instead of attribute can address this limitation.

  https://github.com/yorkie-team/yorkie-js-sdk/blob/2f067ba23994c10ce8791109e0bd915f2e05e62d/test/integration/text_test.ts#L683-L689

3. [ex8] Inheritance of the previous character's style when inserting text at a span boundary

• Description of test case:
  

• In Yorkie, the previous character's style is not inherited.
  https://github.com/yorkie-team/yorkie-js-sdk/blob/2f067ba23994c10ce8791109e0bd915f2e05e62d/test/integration/text_test.ts#L726-L733

• However, this works in the Quill example since the Quill editor manages style inheritance and then applies it to Yorkie.
  

Why: For preserving user intent in rich-text editing