AtlasEngine: Block chars done right-er

[lhecker]

This commit makes the following improvements:

• Only adjust block characters that come from fallback fonts. This ensures
  that the glyphs of the chosen font all look exactly as they were designed.
• When adjusting the size, use the fallback font's full block glyph U+2588
  to determine the size that the given glyph should have.

Closes #14098

Validation Steps Performed

• Print UTF-8-demo.txt in Consolas.
• All block glyphs look uniform. ✅

[lhecker]

before:

after:

[DHowett]

I once again cannot profess to understand the math changes without scrutinizing them, but I will "go with it"

[ghost]

Hello @DHowett!

Because this pull request has the AutoMerge label, I will be glad to assist with helping to merge this pull request once all check-in policies pass.

p.s. you can customize the way I help with merging this pull request, such as holding this pull request until a specific person approves. Simply @mention me (@msftbot) and give me an instruction to get started! Learn more here.

[lhecker]

I once again cannot profess to understand the math changes without scrutinizing them, but I will "go with it"

The DWRITE_OVERHANG_METRICS math is difficult to understand. I wrote it by picking random fonts and sizes and glyphs and painstakingly comparing my own calculations against what GetOverhangMetrics returns until it all matched up perfectly up to the last float digit. To be honest, I don't think there's any comment that can sufficiently explain that code - it simply is confusing, because it is confusing.

If some other code is difficult to understand let me know and I'll try to add some better comments!
BTW the _r.fontFaces matching is extremely bodgy and depends on the existence of the so called ActiveFontFace class (you can look it up internally). I asked the DWrite folks if we can depend on it, but I haven't heard back yet. Basically, my argument for depending on bodgy behavior is because the new-new AtlasEngine needs to hash font-face/glyph-index tuples and you can't do that if you can't hash font faces. Windows 11 22H1 now got a IDWriteFontFace5::Equals method, but as I put it: "you can't hash an equals() method". ActiveFaceCache on the other hand has continuously existed since 2011.

[ghost]

🎉Windows Terminal Preview v1.16.3463.0 and v1.16.3464.0 has been released which incorporates this pull request.:tada:

Handy links:

• Release Notes
• Store Download