ENH: speed up _harmonize_tensor

Common case is just a single element, thus that the overhead of NumPy is
very large.

python/interpret-core/interpret/glassbox/_ebm/_merge_ebms.py

@@ -5,7 +5,7 @@
 import warnings
 from collections.abc import Sequence
 from itertools import chain, count
-from math import isnan
+from math import isnan, prod
 from typing import List, Type
 
 import numpy as np
@@ -257,7 +257,7 @@ def _harmonize_tensor(
             map_bins[bin_idx]
             for map_bins, bin_idx in zip(mapping, old_reversed_bin_idxs)
         ]
-        n_cells2 = np.prod([len(x) for x in cell_map])
+        n_cells2 = prod(map(len, cell_map))
         val = 0 if n_multiclasses == 1 else np.zeros(n_multiclasses, np.float64)
         total_weight = 0.0
         for cell2_idx in range(n_cells2):