fix(traceback): stackmap 的文件名全部退化为 ?? —— _tb_parse_dec 不跳前导空格

dwarfdump 的方括号索引右对齐（file_names[  1] / include_directories[ 12]），_tb_bracket_idx 把 '  1]' 交给 _tb_parse_dec，首字符非数字使循环体一次不进 -> 全部 pidx=0；行记录 fidx=1..N 永不匹配 -> 每条记录落到 '??'（实证 9146/9146 条共用同一个 '??' 串，而函数名/行号均正确）

_tb_parse_dec 增跳前导空格与制表符（修复文件名与目录索引两处）

dir_index: 偏移 st+9 指向冒号本身，恒解析为 0 -> 相对文件名拼不出绝对路径；改为 st+10

App/lib/Phase/BuildPipeline.vp

@@ -1403,8 +1403,21 @@ def _tb_join(mb: memhub.MemBuddy*, base: str, name: str) -> str:
 
 
 def _tb_parse_dec(s: str) -> int:
+    """Parse a leading decimal integer, skipping leading blanks.
+
+    dwarfdump right-aligns the bracketed indices to a 3-wide field
+    (`file_names[  1]:` / `include_directories[ 12]`), so `_tb_bracket_idx`
+    hands us a string whose first char is a space. Without the blank skip the
+    digit loop never runs and every index parses as 0 -> every file record's
+    pidx becomes 0 while a row's fidx is the real 1..N fit -> `_tb_file_rec`
+    never matches and EVERY row degrades to the "??" placeholder. Measured on
+    Test/LLvmLiteTest: all 9146 stackmap records carried the single '??' file
+    string while function names and line numbers were correct.
+    """
     v: int = 0
     i: sizet = 0
+    while _tb_cc(s, i) == 32 or _tb_cc(s, i) == 9:
+        i += 1
     while _tb_cc(s, i) >= 48 and _tb_cc(s, i) <= 57:
         v = v * 10 + (_tb_cc(s, i) - 48)
         i += 1
@@ -1834,7 +1847,9 @@ def _tb_parse_dump(mb: memhub.MemBuddy*, dump: str) -> TbDump | t.CPtr:
                     pkind = 0
                 continue
             if _tb_startswith(st, "dir_index:") != 0 and pkind == 1:
-                pdir = _tb_parse_dec(_tb_ltrim(st + 9))
+                # `"dir_index:"` 含冒号共 10 字符：st + 9 指向冒号本身，解析恒得 0
+                # （目录索引全丢 → 相对文件名拼不出绝对路径，File 行只剩裸文件名）。
+                pdir = _tb_parse_dec(_tb_ltrim(st + 10))
                 continue
             if _tb_startswith(st, "Address") != 0 and _tb_contains(st, "Column") != 0:
                 if pkind == 1 and pname is not None and file_count < TB_FILE_CAP: