Better translations for UI copy

[Onurbon]

The UI is currently translated automatically using LLMs as part of the translation step:

talk-to-the-city-reports/pipeline/steps/translation.py

Line 33 in 0debc1a

UI_copy = ["Argument", "Original comment", "Representative arguments",

It would however be better to improve the copy by hardcoding the translations.
We could start by providing UI copy in Traditional Chinese and then re-generate tttc.dev/genai

[mashbean]

Is there any default prompt during translation? I may provide the testing prompt when translating to ZH-TW and try if it is better!