[TextServer] Fix get_word_breaks and its uses.

[bruvzg]

Fixes #78646

[YuriSizov]

@Rindbee Could you test these changes? Also maybe @timothyqiu could help?

@bruvzg And could you rebase your PR so it runs fresh CI, since its been too long? 🙂

[Rindbee]

The stuttering of tts is gone. but its smoothness is not as good as using UBRK_SENTENCE instead of UBRK_LINE. Maybe when used for tts, special parameters or methods can be provided.

The results returned by string_get_word_breaks are still not particularly ideal.

English test text: "Godot Engine, the game engine you've been waiting for."

Before:
chars_per_line = 0:  [0, 5, 6, 12, 14, 17, 18, 22, 23, 29, 30, 33, 34, 36, 37, 41, 42, 49, 50, 54] 

[Godot]
[Engine]
[the]
[game]
[engine]
[you]
[ve]
[been]
[waiting]
[for.]

chars_per_line = 5:  [0, 5, 6, 11, 11, 12, 14, 17, 18, 22, 23, 28, 28, 33, 34, 36, 37, 41, 42, 47, 47, 49, 50, 54]

[Godot]
[Engin]
[e]
[the]
[game]
[engin]
[e you]
[ve]
[been]
[waiti]
[ng]
[for.]

After:
chars_per_line = 0:  [0, 5, 6, 12, 14, 17, 18, 22, 23, 29, 30, 33, 34, 36, 37, 41, 42, 49, 50, 53] 

[Godot]
[Engine]
[the]
[game]
[engine]
[you]
[ve]
[been]
[waiting]
[for]

chars_per_line = 5:  [0, 5, 6, 11, 11, 13, 14, 17, 18, 22, 23, 28, 28, 33, 33, 36, 37, 41, 42, 47, 47, 49, 50, 54]

[Godot]
[Engin]
[e,]
[the]
[game]
[engin]
[e you]
['ve]
[been]
[waiti]
[ng]
[for.]

It is worth noting that the test results for the English text changed.

Chinese test text: "天行健，君子以自强不息；地势坤，君子以厚德载物"

Before:
chars_per_line = 0:  [0, 1, 0, 1, 0, 3, -1, 0, -1, 0, -1, 0, -1, 0, -1, 0, -1, 0, 10, 11, -1, 0, -1, 0, 14, 15, -1, 0, -1, 0, -1, 0, -1, 0, -1, 0, -1, 0, 22, 23] 

[天]
[天]
[天行健]
[]
[]
[]
[]
[]
[]
[息]
[]
[]
[坤]
[]
[]
[]
[]
[]
[]
[物]

chars_per_line = 5:  [0, 3, 10, 15, 22, 23]

[天行健]
[息；地势坤]
[物]

After:
chars_per_line = 0:  [0, 1, 1, 2, 2, 3, 4, 6, 6, 7, 7, 9, 9, 11, 12, 14, 14, 15, 16, 18, 18, 19, 19, 21, 21, 23] 

[天]
[行]
[健]
[君子]
[以]
[自强]
[不息]
[地势]
[坤]
[君子]
[以]
[厚德]
[载物]

chars_per_line = 5:  [0, 4, 4, 9, 9, 14, 14, 19, 19, 23]

[天行健，]
[君子以自强]
[不息；地势]
[坤，君子以]
[厚德载物]

When chars_per_line = 0, the breaks here seem to have no rules.
In the case of chars_per_line = 5, if a punctuation mark is encountered, English will still break even if it does not reach 5, but Chinese will not.

[bruvzg]

The results returned by string_get_word_breaks are still not particularly ideal.

Not sure if I see anything wrong with it, the difference is only in presence of space, English text breaks on punctuation since it can't fit the next word and a space, so behavior is the same for both.

[akien-mga]

Thanks!