Unicode characters mess up text shadow

Partner Rank: 3

Verification builds: 1.14.0.30, 1.2, 1.16.0.51
Regression builds: None, this bug existed back in 1.2

Summary: When a message contains any "special" unicode character, the text shadow offset is calculated wrongly causing the shadow to be misaligned.

Impact: For two or more upcoming projects, this makes it difficult to read voiceover subtitles.

Repro Steps:
1. Look down at a white block like snow
2.

/titleraw @a title {"rawtext":[{"text":"test message"}]} 

3.

/titleraw @a title {"rawtext":[{"text":"\ue000 test message"}]}

Expected Results: Text shadow looks correct for both messages

Observed Results: Text shadow is only correct for first message

Screenshots/Videos attached: Yes