Avoid wrapping a space onto its own newline

[grovesNL]

Fixes #155

There was some behavior added in #78 that intentionally seems to break trailing spaces in this way, but it doesn't seem be necessary in the test cases I've tried (mostly just experimenting with the rich-text example).

If anyone has a test case that this regresses, please let me know! As far as I can tell this is an improvement for typical cases.

[grovesNL]

cargo-deny failure is unrelated - caused by multiple version of syn and bitflags

[grovesNL]

It seems to cause test_english_mixed_with_arabic_paragraph_rendering to fail - looking into this now

Edit: the test output changed from this

Before

After (more space after render)

It looks like the space might be wider than expected so this might be a regression, but I don't think that was the intent of the original code path.

[jackpot51]

@grovesNL unfortunately that does look like a regression. Hopefully it is not too hard to fix.

[alokedesai]

@grovesNL lmk if there's anyway I can help get this PR past the finish line.

I just spent quite a few quite a few hours debugging a text layout issue in Warp that ultimately seems to be caused by this. Would love to see this fixed and happy to lend a hand if it would useful.

[grovesNL]

@alokedesai absolutely, any help looking into this would be great! I was planning to step through the positioning in a debugger to see how we end up with two spaces there but might not have time until another week or so.

Based on the test case failure it seems like there's some kind of unexpected interaction when we have left-to-right and right-to-left text spans that end up on the same visual line.

[grovesNL]

#226 fixes this without the regression