Calculate the Huffman codebook without MP_QSTRs

[ciscorn]

The Huffman codebooks for the translations are currently computed with the translated texts and MP_QSTRs. However, I don't think they are actually used for QSTRs.

We can decrease the entropy a bit by calculating the codebook based only on the translation texts.

BOARD=feather_m0_express TRANSLATION=ja

(before) // 9342 bytes worth of translations compressed
↓↓↓
(after)  // 9149 bytes worth of translations compressed

BOARD=feather_m0_express

• en: 8398 bytes → 8239 bytes
• es: 9617 bytes → 9561 bytes
• ko: 8815 bytes → 8651 bytes

[jepler]

Excellent find, thank you !

[tannewt]

Thank you! I intended on compressing QSTRs too but never did.