No save dialog if another widget shares identical model

[colin-grant-work]

What it does

Fixes #9474 by modifying Saveable.apply to accept a function to generate a list of widgets that may share the same underlying saveable resource as the widget being closed. If any widget does, then the widget is closed without saving and without reverting the saveable's state.

How to test

1. Open two editors for the same file.
2. Dirty the file.
3. Close one editor.
4. Observe that no dialog is shown asking whether you would like to save and that the other editor remains dirty.
5. Close the other editor.
6. Observe that a dialog is shown asking whether you would like to save.

---

The tricky part is the multi-close commands, close all, close others, and close to the right.

1. Use one of the above named command such that you would not close all instances of a document open in multiple editors.
2. You should not be prompted to save.
3. Use the same command in such a way that you would close all instances
4. You should be prompted to save, and if you cancel, nothing should be closed.

Review checklist

• as an author, I have thoroughly tested my changes and carefully followed the review guidelines

Reminder for reviewers

• as a reviewer, I agree to behave in accordance with the review guidelines

[msujew]

The code looks code to me, and the functionality of this change as well:

• Opening multiple editors for the same model (splitting) continues to work correctly
• Closing a dirty editor whose model is shared with another widget does not prompt a confirmation message
• Using the Close All context menu entry closes all duplicated editor models of the same group, exactly replicating the behavior of vscode
• Closing the last instance of a dirty editor model prompts a confirmation message, as well as closing the app.

[colin-grant-work]

Thanks, @msujew. I'll merge this after the release.

[colin-grant-work]

@vince-fugnitto, I've disabled the run reference code lens test with a TODO. Does that seem reasonable?

[paul-marechal]

@colin-grant-work please consider merging/squashing the following instead: #10688

This way you can squash and merge the current PR without the noise from skipping that failing test.

[paul-marechal]

@colin-grant-work you can now remove 52da73d as #10688 got merged :)