Remove long_to_double from MouseEvent API

[queengooborg]

This PR removes the various long_to_double entries from the MouseEvent API. Overall, the difference between a long and a double is more floating point accuracy, but I can't imagine a scenario where a web developer will require such high levels of precision, not to mention this is near impossible to test, and only really observable in the WebIDL.

[foolip]

These entries are very poorly described, but are about this change I think:
https://chromestatus.com/feature/5117956774166528

(The versions don't match, so I'm not sure...)

https://chromestatus.com/feature/6169687914184704 is a similar change.

[queengooborg]

I'm now realizing that I was seriously confusing long with float, and was thinking that long was a decimal format...oops. As such, I think we should keep it, but maybe give it a better name? (And maybe de-duplicate the entry, so we only have one long_to_double?)

[foolip]

Yeah, I think a single subfeature per interface where some members changed from always integer to sometimes floating point would be good. "fractional coordinates" is actually a pretty good description I think.

I see that https://chromestatus.com/feature/6169687914184704 says:

only browser-fired events that will change are PointerEvents. MouseEvents will continue to have integral coordinates for backward compatibility

If that's still true, then these entries on MouseEvent seems kind of pointless.

[foolip]

This all originates from the wiki table:
https://web.archive.org/web/20180421035120/https://developer.mozilla.org/en-US/docs/Web/API/MouseEvent#Browser_compatibility

I'm pretty sure that the single "MouseEvent.screenX, MouseEvent.screenY, MouseEvent.clientX, and MouseEvent.Y are double instead of long." feature there was based on https://chromestatus.com/feature/6169687914184704 and was then turned into multiple features when it was ported to BCD in #1883.

I've just tested in Chrome 91 to confirm that MouseEvent coordinates are still integers, while PointerEvent has fractional coordinates. In fact, even though MouseEvent does have some double members in Chromium's IDL, it doesn't seem possible to actually observe that, because those same getters floor the returned value to integers:
https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/core/events/mouse_event.h;l=138-146;drc=c629498025fa2b88f251047c22b06ca840c05f5e

A further nail in the coffin is that the spec uses long for these members:
https://w3c.github.io/uievents/#idl-mouseevent

[foolip]

@ddbeck let me know if you need any more info for the release notes here, or if it seems suspect when you try to explain it.

[foolip]

Note also that the real change/feature is already represented here:
https://developer.mozilla.org/en-US/docs/Web/API/PointerEvent/pointerType#browser_compatibility

[foolip]

And I've sent #11327 to link bugs for that entry.