fix: handle closing webContents in BrowserViews

[codebytere]

Description of Change

Closes #37356.
Closes #37419.
Closes #37378.
Refs #37205.

Take 2.

The combination of #35509 and #35603 meant that if a user called webContents.destroy() on a BrowserView's webContents, the associated DraggableRegionProvider would not be removed from the NativeWindow, as the only codepath for doing so relied on the user removing the entire BrowserView via BrowserWindow.removeBrowserView. The first approach I tried fixed a symptom, but then prevented the destroyed event from emitting if a BrowserView still existed.

This PR undoes that change in favor of this better change, which ensures the DraggableRegionProvider isn't left around for a ghost BrowserView webContents.

Tested with:

• https://gist.github.com/YauheniBH-EF/190ea8ff42c09a15763705ed4647b02e to prevent regression of [Bug]: window.close() destroys all child BrowserView-s even if window's 'beforeunload' handler has 'e.returnValue = false' #37162
• https://gist.github.com/mgalla10/cdff38e750ae49d01500445b019493ca

Checklist

• PR description included and stakeholders cc'd
• npm test passes
• tests are changed or added
• PR release notes describe the change in a way relevant to app developers, and are capitalized, punctuated, and past tense.

Release Notes

Notes: Fixed destroyed event not emitted on close for BrowserView.webContents.

[nornagon]

any chance we could get a test? esp, window.close() from the web content not actually closing the BrowserView seems like something we could have a regression test for!

[release-clerk]

Release Notes Persisted

Fixed destroyed event not emitted on close for BrowserView.webContents.

[trop]

I was unable to backport this PR to "22-x-y" cleanly;
you will need to perform this backport manually.

[trop]

I have automatically backported this PR to "24-x-y", please check out #37450

[trop]

I have automatically backported this PR to "23-x-y", please check out #37451

[pushkin-]

When can this be backported to v22?

[wujinli]

When can this be backported to v22?

As #37205 says, the crash is caused by the NativeWindow::NonClientHitTest call. the bug in "NativeWindow::NonClientHitTest" is introduced in #36230. However, the branch of 22-x-y does not contain this refactor.
In my tests, it didn't crash even with revert #37205.
Note: The test is the same as #37205 introduced.

So, just revert it(#37205) in 22-x-y