Fix Japanese vocabulary in SharedRules (general.yaml, default.yaml)

[yasumorishima]

Continues the Japanese rule clean-up. This one is the shared rules, so every
one of these words is spoken by both ClearSpeak and SimpleSpeak.

SharedRules/general.yaml

English	ja before	reads as	now
vector (7 places)	コンテンツ	"content"	ベクトル
end of matrix/vector/determinant (5)	リリース	"release"	行列終了 / ベクトル終了 / 行列式終了
by 1 column (2)	1列ずつ	"one column at a time"	かける 1 列
N, the letter in ℕ²	ネクタイ	"necktie"	エヌ
liquid / gas / aqueous	液体液体 / ガスレンジ / アキュース	doubled / "gas cooker" / not a word	液体 / 気体 / 水溶液
quadruple bond	四倍の結束	"quadruple solidarity"	四重結合

definitions.yaml already reads vector as ベクトル and ℕ as エヌ, and the
matrix branch beside the vector one was already 行列, so these were single
branches left behind.

For end, the file already had 行列終了 for the combined "end matrix"
(line 683), so the noun branches now carry the whole phrase and the separate
end token is emptied — the same shape the file already uses for the English
article ("the" → t: "").

The three sibling bonds were transliterations (シングルボンド / ダブルボンド /
トリプルボンド), so all four now use the terms taught in Japanese chemistry:
単結合 / 二重結合 / 三重結合 / 四重結合.

Both files

sub was サブサブ, the word "sub" doubled, and super was スーパー, which in
Japanese means a supermarket. The verbose branch beside each of them already
said 下付き文字 / 上付き文字, so the short branches are now 下付き / 上付き.

Their end markers were either English word order or transliterations
(終了サブスクリプト, エンドサブ, エンドスーパー, 終了スーパースクリプト).
Japanese puts the noun first, which ClearSpeak_Rules.yaml already does with
上付き終了 / 根号終了 / 分数終了, so they are now 下付き終了 / 下付き文字終了 /
上付き終了 / 上付き文字終了.

default.yaml also had end → リリース for the end of a fraction, which is
only spoken to blind readers; it is now 分数終了.

Every occurrence of each of these words in Rules/Languages/ja/SharedRules/
is fixed, so the two files agree with each other.

Checks

• audit-translations ja: untranslated text 3577 → 3493. The drop is exactly
  84, the number of lines promoted from t: to T:; missing rules 0, extra
  rules 0, and rule differences unchanged at 28.
• Adds ja tests for the number-set letters, the vector / matrix /
  determinant end markers, the short sub and superscript readings, and the
  chemical states and bonds.

One thing I did not change

In default-column-matrix, the opening branch says vector under
ClearSpeak_Matrix = 'EndVector' but the closing branch says end matrix
unconditionally (general.yaml line 683, en line 681). A one-column vector
with four or more rows therefore opens as a vector and closes as a matrix.
That is the same in English, so it looks like an en-side question rather than
a translation one and I left both alone — happy to follow up if you would
like it changed in en first.

🤖 Generated with Claude Code