fix(transforms): make _to_int_list ROI coercion correct and type-safe

The prior _to_int_list used issequenceiterable, which broke in three ways
once the missing import was restored and the code actually ran:

- str/bytes were no longer rejected (issequenceiterable("10") is False, so
  "10" coerced to the scalar [10] instead of raising, silently building
  wrong slices);
- mypy could not narrow the union through issequenceiterable, so the
  iterate/scalar branches failed type checking.

Rewrite the helper to reject str/bytes explicitly, then coerce via the
existing ensure_tuple helper, which handles scalars, sequences, tensors and
ndarrays uniformly and is mypy-clean. Likewise return the crop slices via
ensure_tuple so compute_slices matches its declared tuple[slice] return
type. Behaviour is covered by test_compute_slices_broadcast.

Signed-off-by: Soumya Snigdha Kundu <soumya_snigdha.kundu@kcl.ac.uk>

Author: aymuos15

monai/transforms/croppad/array.py

@@ -344,9 +344,9 @@ def compute_pad_width(self, spatial_shape: Sequence[int]) -> tuple[tuple[int, in
 
 def _to_int_list(data: Sequence[int] | int | NdarrayOrTensor) -> list[int]:
     """Coerce an ROI spec (scalar, sequence, tensor or ndarray) to a list of Python ints."""
-    if issequenceiterable(data):
-        return [int(i) for i in data]
-    return [int(data)]
+    if isinstance(data, (str, bytes)):
+        raise TypeError("ROI specs must be integers or sequences of integers, not strings.")
+    return [int(i) for i in ensure_tuple(data)]
 
 
 def _broadcast_int_pair(
@@ -412,7 +412,7 @@ def compute_slices(
                 starts, ends = _broadcast_int_pair(roi_start, roi_end)
                 starts = [max(s, 0) for s in starts]
             # clamp each end to its own start so no slice has negative width
-            return tuple(slice(s, max(e, s)) for s, e in zip(starts, ends))
+            return ensure_tuple(slice(s, max(e, s)) for s, e in zip(starts, ends))
 
     def __call__(  # type: ignore[override]
         self, img: torch.Tensor, slices: tuple[slice, ...], lazy: bool | None = None