refactor(includes): vipersimd c.LLVMIR 改 LOut 模式 + 文档同步新语义

vipersimd.vp 16 处 simd_* 数学 intrinsic 改 c.LLVMIR(f"{c.LOut(out)} = call ...")
+ return out 模式；TestFBProject 2 处同步；c.py stub 与 wiki 文档
更新为"无返回值直接嵌入原始 IR 文本"语义。

Test/TestFBProject/App/main.vp

@@ -138,7 +138,8 @@ def main() -> t.CInt | t.CExport:
     testcheck.section("Test LLVMIR: inline LLVM IR add")
     x: t.CInt = 10
     y: t.CInt = 20
-    r: t.CInt = c.LLVMIR(f"add i32 {c.LInp(x)}, {c.LInp(y)}", t.CInt)
+    r: t.CInt
+    c.LLVMIR(f"{c.LOut(r)} = add i32 {c.LInp(x)}, {c.LInp(y)}")
     print("LLVMIR add result:", r)
     testcheck.check(r == 30, "LLVMIR add (10+20=30)", "LLVMIR add result mismatch")
 
@@ -152,7 +153,7 @@ def main() -> t.CInt | t.CExport:
 
     testcheck.section("Test LLVMIR with LOut")
     out_val: t.CInt = 0
-    c.LLVMIR(f"{c.LOut(out_val)} = add i32 {c.LInp(x)}, {c.LInp(y)}", t.CInt)
+    c.LLVMIR(f"{c.LOut(out_val)} = add i32 {c.LInp(x)}, {c.LInp(y)}")
     print("LLVMIR LOut result:", out_val)
     testcheck.check(out_val == 30, "LLVMIR LOut (10+20=30)", "LLVMIR LOut result mismatch")