Several attributes and parameters were typed narrower than what the
code actually stores or accepts, silently turning defensive branches
into dead code:

- Macro.__init__ had a bytes-decode branch that could never run since
  the preceding coding-declaration loop already crashes on bytes input
  (regex matching a str pattern against bytes lines); removed the
  dead branch instead of half-fixing bytes support nothing exercises.
- ProfileDir.copy_config_file had a `path is None` fallback that no
  caller ever hits, since `path: Path` is never optional; removed it.
- get_line_number_of_frame checked `co_filename is None`, but
  typeshed (correctly) types co_filename as always `str`; dropped the
  dead branch and its stray debug print.
- FrameInfo._sd is set from stack_data.FrameInfo.stack_data() results,
  which can be RepeatedFrames or None, not just FrameInfo as declared;
  widened the attribute type and scoped the resulting union-attr
  errors to where callers already guard against RepeatedFrames.
- InteractiveShell.data_pub_class and .pylab_gui_select were declared
  as plain `= None` attributes (inferred type `None`), making every
  branch that actually uses them unreachable; gave them real types.
- module_paths.py and PyColorize.py had comparisons/branches that are
  genuinely unreachable per their stubs/logic but intentional at
  runtime (loader-is-a-meta_path-finder check, style property kept
  only to satisfy its return type); scoped-ignored with a comment.
- async_helpers._asyncio_event_loop lacked a type annotation, so mypy
  inferred it as always-None and flagged the "or" short-circuit as
  dead; annotated it as Optional.
- ultratb.py's structured_traceback methods handle `etb` being a
  tuple (chained-exception recursion) or a list, and one handles
  `etype` being None (via `structured_traceback(*sys.exc_info())`),
  none of which fit the public `TracebackType | None` / `type`
  signatures; scoped-ignored with comments rather than widening the
  signature across the whole TB class hierarchy.