Rich Text: Remove updateContent function

[aduth]

This pull request seeks to simplify RichText to have a single function for setting content, vs. the current updateContent and setContent, by removing updateContent.

It's current implementation includes three notable differences from just calling setContent directly, addressed below:

• Prevent the editor from adding an undo level. As I remarked at Proposal: history "buffer"/overwrite, sync RichText history records #4956 (comment), from what I can tell, there does not appear to be an undo level being added if this were not present.
• Restoring bookmark selection position. This seems to trace back to its very first implementation from Editable: implement TinyMCE on the core Editable component #330. It's not clear to me however what we're trying to achieve here (preserving caret location? when does this matter?). It's also the cause of bugs like the one described in Inline boundary and arrow navigation edge case #5095 (comment) , where focus jumps unpredictably when updateContent is called on a block split.
• Setting this.savedContent to the current value. This should take place in setContent if we're assuming that this value is always to be kept in sync with the current content of the editor instance.

Testing instructions:

Verify that there are no regressions in the behavior of the editor, particularly when splitting content, which until the merge of #7482 causes updateContent to be called.

[youknowriad]

• Dropping the focus management breaks the "merge". The focus is not restored at the right position when merging two paragraphs.

• I believe the undo behavior is also necessary in this case or in cases the value of a RichText is updated programmatically (the change is not triggered by the RichText element itself)

[aduth]

But all tests have passed successfully!

😉

[youknowriad]

hahaha :) we definitely need a test for that. I guess we only test the content and not the cursor position.

[aduth]

Added a failing test in dd0e347.

[aduth]

I added back the selection setting, this time with the condition that the editor has focus when content is being set. This is in combination with a new end-to-end test which would fail without the condition (as it does in master as well, even including a ZWSP in the saved content).

See 01e0501.

[ellatrix]

Looks good to me! The only thing I could find is weird undo records on split, but that's in master too.