Use np.intp for 32-bit compatibility (#410) (#485)

Fixes a test failure on 32-bit platforms (e.g. i386) due to NumPy
refusing to safely cast between int64 and int32 in the .repeat() call.

Casting the number of intervals to np.intp ensures compatibility with
the platform's expected integer size for indexing operations, preventing
the TypeError observed on i386.

Author: EdwardBetts

exchange_calendars/calendar_helpers.py

@@ -575,7 +575,7 @@ def _create_index_for_sessions(
 
         # evaluate number of indices for each session
         num_intervals = (end_nanos - start_nanos) / self.interval_nanos
-        num_indices = np.ceil(num_intervals).astype("int64")
+        num_indices = np.ceil(num_intervals).astype(np.intp)
 
         if force_close:
             if self.closed_right: