Improve Viewport and Viewport::WalkInBounds

[lhecker]

This removes all of the 2D iteration machinery. Imagine the text buffer as a Cell[w][h] grid. Clearly, this is identical to a Cell[w*h] array, which shows that copying between overlapping ranges only needs either forward or backward copying, and not left/right/top/down.

With WalkDir removed, WalkInBounds can be rewritten with basic arithmetic which allows pos to be an exclusive end coordinate.

[lhecker]

Section 11.3.6.8 DECCRA seems okay in vttest:

Not sure if there's anything else I can do to test the DetermineWalkDirection function.

[zadjii-msft]

I'd be reluctant to merge this for 1.21 based solely on the fact that these are used everywhere and I'm guessing that bugs in the delta will be very subtle. But I'd be very happy to merge early in 1.22 when we've got more runway to validate.

(I'm also open to being overruled)

[j4james]

FWIW, I've run a number of my own rectangle and scrolling tests on this branch, and I couldn't see any regressions.

[carlos-zamora]

@lhecker Make sure you test this with Narrator. The bounding rects aren't updating for me.

[lhecker]

@carlos-zamora I believe this is a race condition in the UIA implementation. I found that it also reproduces in the stable builds (here: v1.20; repeatedly open cmd and hold any key immediately):

The new-ish Terminal::_assertLocked() catches this bug: The culprit is the constructor which doesn't lock the console when asking for the viewport and cursor position.

[DHowett]

I was scared too, can't figure out why