fix: convert wa-row voiced kana (U+30F8/U+30F9) in toHiraganaCase and toKatakanaCase

[greymoth-jp]

toHiraganaCase and toKatakanaCase already handle two of the four wa-row voiced katakana: ヷ (U+30F7) is decomposed to わ゛ and ヺ (U+30FA) to を゛, and toKatakanaCase composes them back. The other two, ヸ (U+30F8) and ヹ (U+30F9), were left out, so they pass through unconverted:

Encoding.toHiraganaCase('ヷヸヹヺ');      // 'わ゛ヸヹを゛'  (ヸ/ヹ unchanged)
Encoding.toKatakanaCase('わ゛ゐ゛ゑ゛を゛'); // 'ヷヰ゛ヱ゛ヺ'  (ゐ゛/ゑ゛ not composed)

The base letters ヰ/ヱ already convert to ゐ/ゑ through the usual katakana/hiragana offset, so the voiced forms now follow the same pattern as ヷ/ヺ:

Encoding.toHiraganaCase('ヷヸヹヺ');      // 'わ゛ゐ゛ゑ゛を゛'
Encoding.toKatakanaCase('わ゛ゐ゛ゑ゛を゛'); // 'ヷヸヹヺ'

Added a test covering the round-trip for the full wa-row voiced family. npm test passes.

[polygonplanet]

@greymoth-jp
Thank you for the thoughtful PR!
I didn't realize that ヸ (U+30F8) and ヹ (U+30F9) were missing. The tests are thorough too, which is a big help. I will go ahead and merge this. Thank you!