fix(rfi): fix edge case where bounds are empty slice.

There is an edge case where slice(None) is passed to mad_cut_rolling
which would end up comparing slice.start (None) to a value, raising an
error. This commit tweaks the implementation to get the same behaviour
without the risk of this error.

ch_util/rfi.py

@@ -229,7 +229,7 @@ def number_deviations(
         arr = auto_vis[:, ind].allgather() if parallel else auto_vis[:, ind]
         # Use NaNs to ignore previously flagged data when computing the MAD
         arr = np.where(flg, arr.real, np.nan)
-        local_bounds = auto_vis.local_bounds if parallel else Ellipsis
+        local_bounds = auto_vis.local_bounds if parallel else slice(None)
         # Apply RFI flagger
         if rolling:
             # Limit bounds to the local portion of the array
@@ -568,7 +568,7 @@ def mad_cut_rolling(
     threshold=5.0,
     freq_flat=True,
     mask=True,
-    limit_range: slice = Ellipsis,
+    limit_range: slice = slice(None),
 ):
     """Mask out RFI by placing a cut on the absolute deviation.
     Compared to `mad_cut_2d`, this function calculates
@@ -622,7 +622,7 @@ def mad_cut_rolling(
     exp_data = np.full(eshp, np.nan, dtype=data.dtype)
     exp_data[foff : foff + nfreq, toff : toff + ntime] = data
 
-    if limit_range is not Ellipsis:
+    if limit_range != slice(None):
         # Get only desired slice
         expsl = slice(
             max(limit_range.start, 0),