perf(prof,gen): 编译热点优化 + 内存池 OOM 诊断接线 + REnum 指针载荷修正

CallGenBase: 模块返回类型文本表改单趟登记+查找，find_func_in_module 合并为单趟链表遍历（PROF 热点）

BuildPipeline: ensure_dir 增加快速路径（已存在目录直接返回）

AssignGenMixin/Translator: alloc 失败打印 Pool/PermPool 统计与阶段上下文（SetAllocCtx）

CallGenCtor: REnum 指针载荷补显式 build_ptrtoint 再 coerce/store；新增 DebugInfoGen.vp

其余为历史累积改动一并入库

App/lib/Phase/Phase1/Phase1Main.vp

@@ -27,6 +27,7 @@ import lib.Phase.Phase1.Phase1Topo as Phase1Topo
 import lib.Phase.Phase1 as Phase1
 
 import lib.core.Gen.GenApi as GenApi
+import lib.core.Gen.DebugInfoGen as DebugInfoGen
 
 
 # ============================================================
@@ -1027,6 +1028,11 @@ def RunPhase1(mb: memhub.MemBuddy*, includes_dir: str, temp_dir: str,
         tr.ModuleSha1 = sha1
         tr.CurrentPackage = GenApi.compute_package_from_relpath(mb, entry.RelPath)
         GenApi.set_current_file_type(file_path)
+        # --traceback: mirror the DWARF switch into the Gen-layer emitter (the
+        # same BuildPipeline._traceback_mode that gates -frame-pointer=all /
+        # dbghelp / the Install() ctor). include/*.vp modules are compiled here,
+        # so this is what gives their frames DWARF line info.
+        DebugInfoGen.SetEnabled(BuildPipeline._traceback_mode)
         GenApi.set_current_file_imports(file_path)
         GenApi.set_current_module_sha1(sha1)
         GenApi.clear_cdefine_constants()
@@ -1058,6 +1064,13 @@ def RunPhase1(mb: memhub.MemBuddy*, includes_dir: str, temp_dir: str,
             failed += 1
             continue
 
+        # --traceback: inject the startup Install() ctor when translating traceback.vp
+        # (its stub/text.ll below are what Phase2 reuses to build the .obj). The
+        # injector lives in the Gen layer (already depends on llvmlite), so this
+        # module gains no cross-module llvmlite symbol dependency.
+        if BuildPipeline._traceback_mode != 0 and BuildPipeline._tb_basename_is_traceback(file_path) != 0:
+            HandlesTranslator.InjectTracebackInstallCtor(tr)
+
         # 诊断：pre_rc 哨兵在 translate 前、覆盖不到 translate 期间的 AST/IR 写穿，故 dump 前立即检查存活块 header/canary
         # （rc=6 块尾 canary 损坏 = 越界写穿数据区），把「dump 时链表成环卡死」变成「定位到 translate 期间写穿发生的文件」。
         post_rc: int = mb.check_alive_blocks(mark1)