[RTL] Fix RichTextLabel vertical alignment calculation regression

[AgZ47]

What problem(s) does this PR solve?

• Assessed Critical issue with RichTextLabel vertical_alignment #122374

Additional information

Commit facdd8e added a zero-width space (U+200B) to lines in RichTextLabel::_shape_line() to ensure uniform line height with fallback fonts. However, appending U+200B causes TextServer to append an extra glyph run that inflates text_buf->get_size().y by 3px for standard fonts. This inflated total height causes VERTICAL_ALIGNMENT_BOTTOM to be offset by 3px and VERTICAL_ALIGNMENT_CENTER to be offset by ~1.5px.

I have used AI to help me as a first time contributor, helped me with some of the git commands and in finding the right file locations.

[bruvzg]

This will reintroduce a bunch of other issues this code fixed.

[AgZ47]

Would restricting the 0x200B insertion to only when the line has no text glyphs be the right direction to preserve the fallback font fix without inflating populated lines?

[AgZ47]

I've updated the PR to guard the 0x200B insertion with txt.is_empty():

• Empty / baseline lines: Still receive 0x200B so fallback fonts and blank lines retain their full baseline height (verified with multi-line spacing preserving the expected 22px gap).
• Populated text lines: Skip 0x200B so text_buf metrics are not artificially inflated, restoring pixel-perfect 0px alignment for both Center and Bottom vertical alignments.