fix: replace strided slicing with out= for NumPy 2.x compatibility

NumPy 2.x rejects strided slicing (e.g., array[::2, :]) as the out=
parameter in reduction operations due to stricter dimension validation.

Before:
  np.sum(cohort_is_amp, axis=1, out=count[::2, cohort_index])

After:
  count[::2, cohort_index] = np.sum(cohort_is_amp, axis=1)

Root cause: NumPy 2.x enforces strict dimension checking for out=
parameters in ufunc.reduce operations. Strided views create non-contiguous
arrays that fail this validation with ValueError.

The explicit assignment approach produces identical results and works with
both NumPy 1.26.x and 2.x.

Fixes coverage job failures under NumPy 2.x test matrix.

Author: adilraza99

malariagen_data/anoph/cnv_frq.py

@@ -573,8 +573,8 @@ def _gene_cnv_frequencies_advanced(
             cohort_is_called = np.take(is_called, sample_indices, axis=1)
 
             # compute cohort allele counts
-            np.sum(cohort_is_amp, axis=1, out=count[::2, cohort_index])
-            np.sum(cohort_is_del, axis=1, out=count[1::2, cohort_index])
+            count[::2, cohort_index] = np.sum(cohort_is_amp, axis=1)
+            count[1::2, cohort_index] = np.sum(cohort_is_del, axis=1)
 
             # compute cohort allele numbers
             cohort_n_called = np.sum(cohort_is_called, axis=1)