feat: 适配 LuaJIT 2.1 (luajit_2.1 分支)

- lib/lua 替换为完整 LuaJIT 2.1 仓库(Makefile+src+dynasm, 官方结构)
- 构建系统改造: lib/lua/CMakeLists.txt 改为调用 LuaJIT 自带 make
  (BUILDMODE=static CFLAGS=-fPIC) 产出 libluajit.a; 顶层链接该静态库
  并加 -Wl,-E 导出符号(静态链 LuaJIT 必需)
- 补充 lua.hpp 转发头兼容桥接层 'lua/lua.hpp' include 路径
- 桥接层沿用 5.1 方案: Registry 存 iopath / luaL_tolstring 兜底 /
  typed interop 无 integer 子类型(LUA_VERSION_NUM=501)
- [关键修复] lua_remove 栈越界 bug(各版本共有, LuaJIT 下显性崩溃):
  * unload_lib: ref 记录的是 gettop+1, 实际移除应用 ref-1
  * clean(): 原逐个 remove 循环恒越界且多次移除后索引偏移,
    属死代码(末尾 settop(L,0) 已清栈), 直接删除
  该 bug 在 Lua 5.4 中被隐式容忍, 在 LuaJIT 的 GC finalizer 中
  表现为 lj_gc_finalize_cdata 段错误
- 测试跨版本断言同步(自 5.1.5)
- 验证: GTest 44/44 通过

lib/lua/src/lj_cconv.h

@@ -0,0 +1,71 @@
+/*
+** C type conversions.
+** Copyright (C) 2005-2026 Mike Pall. See Copyright Notice in luajit.h
+*/
+
+#ifndef _LJ_CCONV_H
+#define _LJ_CCONV_H
+
+#include "lj_obj.h"
+#include "lj_ctype.h"
+
+#if LJ_HASFFI
+
+/* Compressed C type index. ORDER CCX. */
+enum {
+  CCX_B,	/* Bool. */
+  CCX_I,	/* Integer. */
+  CCX_F,	/* Floating-point number. */
+  CCX_C,	/* Complex. */
+  CCX_V,	/* Vector. */
+  CCX_P,	/* Pointer. */
+  CCX_A,	/* Refarray. */
+  CCX_S		/* Struct/union. */
+};
+
+/* Convert C type info to compressed C type index. ORDER CT. ORDER CCX. */
+static LJ_AINLINE uint32_t cconv_idx(CTInfo info)
+{
+  uint32_t idx = ((info >> 26) & 15u);  /* Dispatch bits. */
+  lj_assertX(ctype_type(info) <= CT_MAYCONVERT,
+	     "cannot convert ctype %08x", info);
+#if LJ_64
+  idx = ((uint32_t)(U64x(f436fff5,fff7f021) >> 4*idx) & 15u);
+#else
+  idx = (((idx < 8 ? 0xfff7f021u : 0xf436fff5) >> 4*(idx & 7u)) & 15u);
+#endif
+  lj_assertX(idx < 8, "cannot convert ctype %08x", info);
+  return idx;
+}
+
+#define cconv_idx2(dinfo, sinfo) \
+  ((cconv_idx((dinfo)) << 3) + cconv_idx((sinfo)))
+
+#define CCX(dst, src)		((CCX_##dst << 3) + CCX_##src)
+
+/* Conversion flags. */
+#define CCF_CAST	0x00000001u
+#define CCF_FROMTV	0x00000002u
+#define CCF_SAME	0x00000004u
+#define CCF_IGNQUAL	0x00000008u
+
+#define CCF_ARG_SHIFT	8
+#define CCF_ARG(n)	((n) << CCF_ARG_SHIFT)
+#define CCF_GETARG(f)	((f) >> CCF_ARG_SHIFT)
+
+LJ_FUNC int lj_cconv_compatptr(CTState *cts, CType *d, CType *s, CTInfo flags);
+LJ_FUNC void lj_cconv_ct_ct(CTState *cts, CType *d, CType *s,
+			    uint8_t *dp, uint8_t *sp, CTInfo flags);
+LJ_FUNC int lj_cconv_tv_ct(CTState *cts, CType *s, CTypeID sid,
+			   TValue *o, uint8_t *sp);
+LJ_FUNC int lj_cconv_tv_bf(CTState *cts, CType *s, TValue *o, uint8_t *sp);
+LJ_FUNC void lj_cconv_ct_tv(CTState *cts, CType *d,
+			    uint8_t *dp, TValue *o, CTInfo flags);
+LJ_FUNC void lj_cconv_bf_tv(CTState *cts, CType *d, uint8_t *dp, TValue *o);
+LJ_FUNC int lj_cconv_multi_init(CTState *cts, CType *d, TValue *o);
+LJ_FUNC void lj_cconv_ct_init(CTState *cts, CType *d, CTSize sz,
+			      uint8_t *dp, TValue *o, MSize len);
+
+#endif
+
+#endif