BUG: fix dtype of include_initial in cumulative_sum

In `concat([zeros(...), x])` zeros must have the same dtype as `x`.

Author: ev-br

Diff for: array_api_strict/_statistical_functions.py

@@ -44,7 +44,7 @@ def cumulative_sum(
     if include_initial:
         if axis < 0:
             axis += x.ndim
-        x = concat([zeros(x.shape[:axis] + (1,) + x.shape[axis + 1:], dtype=dt), x], axis=axis)
+        x = concat([zeros(x.shape[:axis] + (1,) + x.shape[axis + 1:], dtype=x.dtype), x], axis=axis)
     return Array._new(np.cumsum(x._array, axis=axis, dtype=dtype), device=x.device)