feat: disable google_japanese_input on affix conversions

[atusy]

In the current implementation, the google_japanse_input source returns invalid candidates for affixes (prefix and suffix).

For example, Dai> gives 第＞ and 大＞.

This PR prevents it by conditionally disabling the google_japanse_input source.

In order to implement the feature, I changed the signature of getHenkanResult() and related methods.

- getHenkanResult(type: HenkanType, word: string)
+ getHenkanResult(word: string, type: HenkanType, affix?: AffixType)

Let me know if reviewer prefers type, word, affix or some other signatures.

[atusy]

Maybe we should add more generalized approach.
Otherwise, we need more patches as we implement 単漢字変換, 数値変換, and else.