修复循环信号与事件采样并接入 LSTP 接触定位，补充八路验证及复用实验

相较上一版 Jacobian 确定性复用更新，本次补齐事件边界一致性、结果两侧采样及接触事件定位；保留已有物性复用和组件力学公式。

- 统一 UD00 信号求值与下一事件查询的绝对时间边界，修复循环边界浮点舍入导致的阶段错位、重复或漏报，并覆盖零时长、多阶段及长周期场景。
- 引入原生输出语义 v2：保留规则网格真实时间，补充内部时间事件和状态事件的左邻及事件后采样，按保存时间、状态和离散模式重放结果。
- 两条代码生成路径均发出 LSTP 接触描述，默认定位间隙过零及非负力模式的力截断；仅在接受事件时更新防重复记录，增加 contactEvents 诊断计数。
- 补充 MASS/LSTP 独立事件实验、八路全曲线与驱动阶段配对评估，以及 Amesim 不连续点输出对照和力差定位报告；MASS 新增释放机制仍保留为独立实验。
- 保存局部 probe、context 访问与回退、shadow replay、R288 real skip/typed replay 及阀门数值尾部诊断工具和报告；未证明净收益的实验不启用为生产默认优化。
- 更新原生运行说明和元件建模规范，补充信号边界、输出语义、接触事件和实验依赖回归测试。

验证：五组专项回归共 34 项全部通过；37 个待提交 Python 文件语法检查通过；git diff --cached --check 通过。

tests/manual/valve_tail_diag.c

@@ -0,0 +1,63 @@
+/* Diagnostic worker only: original tail executes on every visit. */
+#include "valve_tail_diag.h"
+#include <windows.h>
+#include <cpuid.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <fenv.h>
+#include <math.h>
+#include <stddef.h>
+int kd_scope;
+uint64_t kd_jac;
+KdMemo kd_memo;
+static KdRecord records[26000];
+static size_t count;
+static uint64_t entries,hits,executions,lookup_ticks,capture_ticks,reset_ticks,tail_ticks;
+static uint64_t base_lookup_ticks,miss_lookup_ticks,hit_lookup_ticks;
+static uint64_t qpc_frequency,tsc_start,qpc_start;
+static FILE *matrix_file;
+static int disabled;
+static void fatal(const char *s){fprintf(stderr,"valve tail: %s\n",s);abort();}
+void kd_eval_exit(double t,const double*y,const double*dy,const double*w,int r,NativePropertyCache*p,NativePipeCache*c,ModelJacobianWorkspace*m){(void)t;(void)y;(void)dy;(void)w;(void)r;(void)p;(void)c;(void)m;}
+void kd_start(void){
+    unsigned a,b,c,d;if(!__get_cpuid(0x80000007,&a,&b,&c,&d) || !(d&(1u<<8)))fatal("invariant TSC required");
+    LARGE_INTEGER f;QueryPerformanceFrequency(&f);qpc_frequency=(uint64_t)f.QuadPart;
+    qpc_start=lp_tick();tsc_start=kd_clock();disabled=getenv("KD_DISABLED")!=NULL;
+    if(getenv("KD_MATRICES")){matrix_file=fopen("jacobians.bin","wb");if(!matrix_file)fatal("matrix file");setvbuf(matrix_file,NULL,_IOFBF,1024*1024);}
+}
+void kd_new_jac(void){uint64_t t=kd_clock();kd_memo.ready=0;reset_ticks+=kd_clock()-t;kd_jac++;}
+void kd_before(KdRecord *r,double p,double T,double pd,double g,double rho){
+    double key[5]={p,T,pd,g,rho};memcpy(r->key,key,40);r->before=kd_env();
+    r->hit=lp_color>=0 && !disabled && kd_lookup(r->key,r->before,&r->cm,&r->vel);
+    r->jac=kd_jac;r->group=lp_color;
+}
+void kd_after(KdRecord *r,double cm,double vel,uint64_t elapsed,int executed){
+    uint64_t capture_start=lp_color<0?kd_clock():0;
+    r->after=kd_env();r->cm=cm;r->vel=vel;r->tail=elapsed;r->executed=executed;r->capture=0;r->pad=0;
+    if(lp_color<0){
+        memcpy(kd_memo.key,r->key,40);kd_memo.cm=cm;kd_memo.vel=vel;kd_memo.env=r->before;
+        kd_memo.ready=kd_finite(r->key) && isfinite(cm) && isfinite(vel) && kd_env_equal(r->before,r->after) && (r->before.cw&63)==63 && (r->before.mxcsr&0x1f80)==0x1f80;
+        r->capture=kd_clock()-capture_start;capture_ticks+=r->capture;
+    }
+    entries++;lookup_ticks+=r->lookup;if(lp_color<0)base_lookup_ticks+=r->lookup;else if(r->hit)hit_lookup_ticks+=r->lookup;else miss_lookup_ticks+=r->lookup;
+    hits+=r->hit;executions+=executed;tail_ticks+=elapsed;
+#if KD_TRACE
+    if(count>=sizeof(records)/sizeof(*records)){fatal("record capacity");}
+    records[count++]=*r;
+#endif
+}
+void kd_matrix(double t,const double*y,const double*m){
+    if(!matrix_file){return;}
+    int saved=errno;fenv_t env;fegetenv(&env);unsigned sse=_mm_getcsr();
+    fwrite(&t,8,1,matrix_file);fwrite(y,8,NSTATES,matrix_file);fwrite(m,8,NSTATES*NSTATES,matrix_file);
+    fesetenv(&env);_mm_setcsr(sse);errno=saved;
+}
+void kd_finish(void){
+    uint64_t t=kd_clock(),q=lp_tick();double frequency=(double)(t-tsc_start)/(q-qpc_start)*qpc_frequency;
+    uint64_t empty[10001];for(int i=0;i<10001;i++){uint64_t a=kd_clock();empty[i]=kd_clock()-a;}
+    FILE *f=fopen("tail-records.bin","wb");if(!f)fatal("records");fwrite(records,sizeof(KdRecord),count,f);fclose(f);
+    f=fopen("empty-clock.bin","wb");fwrite(empty,8,10001,f);fclose(f);
+    f=fopen("tail.json","wb");if(!f)fatal("summary");
+    fprintf(f,"{\"frequency\":%.9f,\"recordBytes\":%llu,\"memoBytes\":%llu,\"jacobians\":%llu,\"entries\":%llu,\"hits\":%llu,\"executions\":%llu,\"lookupTicks\":%llu,\"hitLookupTicks\":%llu,\"missLookupTicks\":%llu,\"baseLookupTicks\":%llu,\"captureTicks\":%llu,\"resetTicks\":%llu,\"tailTicks\":%llu,\"skipEnabled\":%d,\"disabled\":%d}\n",frequency,(unsigned long long)sizeof(KdRecord),(unsigned long long)sizeof(KdMemo),(unsigned long long)kd_jac,(unsigned long long)entries,(unsigned long long)hits,(unsigned long long)executions,(unsigned long long)lookup_ticks,(unsigned long long)hit_lookup_ticks,(unsigned long long)miss_lookup_ticks,(unsigned long long)base_lookup_ticks,(unsigned long long)capture_ticks,(unsigned long long)reset_ticks,(unsigned long long)tail_ticks,0,disabled);fclose(f);
+    if(matrix_file)fclose(matrix_file);
+}