Handle empty variant_query results gracefully instead of raising ValueError

When variant_query filters leave zero SNPs in snp_allele_frequencies_advanced() and
aa_allele_frequencies_advanced(), the code previously raised ValueError. This is
a pre-existing fragility exposed by the region slicing fix (is not None checks
can yield more accurate but smaller variant sets).

Changed both sites to emit a UserWarning and continue, returning an empty but
well-formed xr.Dataset with stable schema. This follows the existing precedent
in aa_allele_frequencies() (line 354) which already handles empty results
gracefully.

Author: khushthecoder

malariagen_data/anoph/snp_frq.py

@@ -671,8 +671,10 @@ def snp_allele_frequencies_advanced(
 
             # Check for no SNPs remaining after applying variant query.
             if np.count_nonzero(loc_variants) == 0:
-                raise ValueError(
-                    f"No SNPs remaining after applying variant query {variant_query!r}."
+                warnings.warn(
+                    f"No SNPs remaining after applying variant query {variant_query!r}. "
+                    "Returning dataset with zero variants.",
+                    stacklevel=2,
                 )
 
             # Convert boolean mask to integer indices for NumPy 2.x compatibility
@@ -813,8 +815,10 @@ def aa_allele_frequencies_advanced(
 
             # Check for no SNPs remaining after applying variant query.
             if np.count_nonzero(loc_variants) == 0:
-                raise ValueError(
-                    f"No SNPs remaining after applying variant query {variant_query!r}."
+                warnings.warn(
+                    f"No SNPs remaining after applying variant query {variant_query!r}. "
+                    "Returning dataset with zero variants.",
+                    stacklevel=2,
                 )
 
             # Convert boolean mask to integer indices for NumPy 2.x compatibility