Improve diagnostics for "Family had N reactants?" kinetics test failure

The dispatch else-branch in kinetics_check_sample_can_react raised a terse
ValueError ("Family had N reactants?: ...") that named neither the family nor
the underlying cause, making CI failures hard to diagnose. This branch is
reached whenever the number of reactant roots that produced at least one usable
sample molecule does not match the family
template, and it raised before the
logging loop ran, discarding any per-sample errors already queued.

Add an explicit check comparing the number of reactant roots that
yielded samples against the family template's reactant count. On a mismatch,
append a descriptive, per-root message  logged together with
the queued per-sample errors (and then raises the standard
"Error Occurred. See log for details."), naming the family and the specific
reactant root(s) with no usable sample.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: rwest

test/database/databaseTest.py

@@ -1577,7 +1577,22 @@ def make_error_message(reactants, message=""):
                 output += "\n" + s.to_adjacency_list(label=s.to_smiles())
             return output
 
-        if len(sample_reactants) == 1 == len(family.forward_template.reactants):
+        expected_reactants = [str(r) for r in family.forward_template.reactants]
+        roots_with_samples = [str(k) for k in sample_reactants.keys()]
+        roots_without_samples = [r for r in expected_reactants if r not in roots_with_samples]
+        if len(sample_reactants) != len(family.forward_template.reactants):
+            # One or more reactant roots produced no usable sample molecule (every candidate was
+            # forbidden by is_molecule_forbidden, or raised UnexpectedChargeError/
+            # ImplicitBenzeneError during make_sample_molecule). Record a descriptive error so it
+            # is logged below alongside any per-sample errors, instead of raising opaquely here.
+            test1.append(
+                f"In family {family_name}, {len(roots_without_samples)} reactant root(s) produced "
+                f"no usable sample molecule: {roots_without_samples}. The family template expects "
+                f"{len(expected_reactants)} reactant(s) {expected_reactants}; only these root(s) "
+                f"yielded samples: {roots_with_samples}. Check the group definitions for the "
+                f"missing reactant root(s)."
+            )
+        elif len(sample_reactants) == 1:
             reactants = list(sample_reactants.values())[0]
             for reactant in reactants:
                 try:
@@ -1670,7 +1685,13 @@ def make_error_message(reactants, message=""):
                     species = rmgpy.species.Species(index=1, molecule=[molecule])
                     species.generate_resonance_structures()
         else:
-            raise ValueError(f"Family had {len(sample_reactants)} reactants?: " f"{', '.join(map(str,sample_reactants.keys())) }")
+            # Reactant count matches the template but is not 1, 2, or 3 (RMG only supports up to
+            # trimolecular). This is not expected for any well-formed family.
+            raise ValueError(
+                f"In family {family_name}, the number of sampled reactant roots "
+                f"({len(sample_reactants)}) matches the template reactant count but is not 1, 2, "
+                f"or 3, which is unexpected: {roots_with_samples}."
+            )
 
         # print out entries skipped from exception we can't currently handle
         if skipped: