Closes https://github.com/ppy/osu/issues/28916.
The previous behaviour *may* have been intended, but it was honestly
quite baffling. This seems like a saner variant.
This sort of thing is bound to happen when rewriting screens from
scratch without invoking abstract eldritch entities sometimes. Damned if
you do, damned if you don't...
Closes https://github.com/ppy/osu/issues/28983.
While the direct cause of this is most likely mouse confine in
full-screen, it shouldn't/can't really be disabled just for this,
and I also get this on linux in *windowed* mode.
In checking other apps, adding some tolerance to this sort of
drag-scroll behaviour seems like a sane UX improvement anyways.
I'm not *super* sure why this works, but it appears to, and my educated
guess as to why is that it counteracts the effects of a change in the SV
of the juice stream by artificially increasing or decreasing the
velocity when running the appropriate path conversions and expected
distance calculations. The actual SV change takes effect on the next
default application, which is triggered by the `Update()` call at the
end of the method.
This is important because the editable path conversions heavily depend
on the value of `JuiceStream.Velocity` being correct. The value is only
guaranteed to be correct after an `ApplyDefaults()` call, which is
triggered by updating the object via `EditorBeatmap`.
If a key is pressed while the pause overlay is visible, the ruleset input manager will not see it, therefore if the user resumes while the key is held then releases the key, the ruleset input manager will not receive the key up event.