fix(输入法): 优化双端联想排序并消除文字刷新闪烁

com.sydxky.rtos.inputmethod/tests/engine.test.mjs

@@ -56,4 +56,14 @@ check(edit('你好',2,'世界',3).limited,'maximum length enforced')
 check(!preview('my-secret',5,true).includes('secret'),'secret preview masked')
 check(chars('😀').length===1,'Unicode length counts codepoints')
 dictionary.clear();check(dictionary.cacheSize===0,'cache release')
+check((await dictionary.lookup('niha',false,{}))[0].word==='你好','unfinished pinyin completes a real word before shorter fallback')
+check((await dictionary.lookup(digitsOf('niha'),true,{}))[0].word==='你好','unfinished T9 shares the completion policy')
+check((await dictionary.lookup(code14Of('niha'),'14',{})).some(x=>x.word==='你好'&&x.consume===4&&x.completion),'14-key completion consumes only the actual input tokens')
+check((await dictionary.lookup('beij',false,{}))[0].word==='北京','word frequency places common completion before rare homophones')
+const continuation=await dictionary.predict('你好')
+check(continuation[0].word==='世界'&&continuation.every(x=>x.consume===0)&&continuation.length<=8,'post-commit continuations use the longest context with bounded results')
+check((await dictionary.predict('你好。')).length===0&&(await dictionary.predict('secret123')).length===0,'punctuation and non-Chinese contexts end prediction')
+check((await dictionary.predict('前文你好'))[0].word==='世界','prediction uses only the bounded suffix before the cursor')
+check((await dictionary.lookup('shijie',false,{}))[0].word==='世界','common full-pinyin homophone is ranked by licensed word frequency')
+check(dictionary.cacheSize<=4,'prediction and completion retain the bounded cache')
 console.log(JSON.stringify({suite:'ime-engine',assertions:count,keys:meta.keys,candidates:meta.candidates,status:'passed'}))