[stantheman0128]

Closes #886.

Problem

Issue #886 and its comments list many STPhrases gaps where shorter MaxMatch entries win over the intended reading. Examples on current packaged dictionaries:

地瓜干 -> 地瓜幹   (want 地瓜乾, same pattern as 葡萄干)
观影后 -> 觀影后   (影后 identity wins; want 觀影後)
控球后卫 -> 控球后衛 (球后 identity wins; want 控球後衛)
年历史 -> 年曆史   (年历 wins; want 年歷史)
天干完 -> 天干完   (天干 wins over 干完; want 天幹完)
得力干将 -> 得力干將 (干将 sword name wins; want 得力幹將)
落地签 / 免签 -> 籤 (character default; want 簽)
纤道 -> 纖道       (want 縴道, same as 纤夫)
并兼任 -> 併兼任   (并兼 wins; want 並兼任)
仲裁制 -> 仲裁製   (裁制 wins; want 仲裁制)
在海里 -> 在海里   (海里 nautical-mile default; want 在海裏)
余华 -> 餘華       (want surname 余華)

Issue body also proposed 干云蔽日 -> 乾雲蔽日. The cited MoE entry (ID 69701) and the 《後漢書》 quote both use 干雲蔽日, and OpenCC already maps 干云蔽日 -> 干雲蔽日. This PR leaves that entry unchanged.

Change

Add longer / clearer STPhrases entries for the reproducible cases above (dried foods, after-compounds, history vs calendar, gan/visa/tow/system/sea/name), plus controls so 影后 / 武后 / 球后 / 封后 / 天干 / 干将莫邪 / 年历 / 裁制 stay as they are.

Dictionary sorted with python data/scripts/sort.py data/dictionary/STPhrases.txt.

Test plan

• New cases BYVoid_OpenCC_Issue_886_* in test/testcases/testcases.json
• Local verification with OpenCC 1.4.1 wheel + branch text dictionaries (see Evidence)
• CI Bazel dictionary / command_line tests (authoritative)

Evidence

Repro before (packaged wheel dictionaries):

s2t: 地瓜干 -> 地瓜幹
s2t: 观影后 -> 觀影后
s2t: 控球后卫 -> 控球后衛
s2t: 年历史 -> 年曆史
s2t: 天干完 -> 天干完
s2t: 得力干将 -> 得力干將
s2t: 落地签 -> 落地籤
s2t: 免签 -> 免籤
s2t: 纤道 -> 纖道
s2t: 并兼任 -> 併兼任
s2t: 仲裁制 -> 仲裁製
s2t: 在海里 -> 在海里
s2t: 余华 -> 餘華
s2t: 干云蔽日 -> 干雲蔽日

After this branch (config JSONs with ocd2 refs switched to text, driving data/dictionary/*.txt on the branch):

s2t: 地瓜干 -> 地瓜乾
s2t: 观影后 -> 觀影後
s2t: 控球后卫 -> 控球後衛
s2t: 年历史 -> 年歷史
s2t: 天干完 -> 天幹完
s2t: 得力干将 -> 得力幹將
s2t: 落地签 -> 落地簽
s2t: 免签 -> 免簽
s2t: 纤道 -> 縴道
s2t: 并兼任 -> 並兼任
s2t: 仲裁制 -> 仲裁制
s2t: 在海里 -> 在海裏
s2t: 余华 -> 余華
s2t: 干云蔽日 -> 干雲蔽日

New #886 cases: 9/9 pass under s2t / s2tw / s2twp / s2hk.

A/B of existing s2t expectations in testcases.json (branch STPhrases vs expectations, excluding the new #886 cases): 93/93 same, 0 diffs.

sort.py idempotence on STPhrases.txt: no diff.

Please treat CI Bazel tests as authoritative; no local C++/Bazel toolchain was available here.

AI assistance

This fix was developed with AI assistance (Cursor / Grok) and verified locally as shown above.