Merge pull request #2941 to tweak SIDT tree generation

iter_item_cap now passed to make_tree_nodes even if cascade algorithm isn't used

tree generation was taking forever because grp.get_extensions has found bond formation extensions that match all the training reactions in the node and are flagged to be further expanded, which explodes the amount of extensions to be evaluated. I see that this issue was addressed in commit b6ae40d, where an iter_item cap will force it to give up the split if it runs into this issue but only when the cascade algorithm is used.

Author: rwest

rmgpy/data/kinetics/family.py

@@ -3294,7 +3294,7 @@ def generate_tree(self, rxns=None, obj=None, thermo_database=None, T=1000.0, npr
             template_rxn_map = self.get_reaction_matches(rxns=rxns, thermo_database=thermo_database, remove_degeneracy=True,
                                                          fix_labels=True, exact_matches_only=True, get_reverse=True)
             self.make_tree_nodes(template_rxn_map=template_rxn_map, obj=obj, T=T, nprocs=nprocs - 1, depth=0,
-                                 min_splitable_entry_num=min_splitable_entry_num, min_rxns_to_spawn=min_rxns_to_spawn,extension_iter_max=extension_iter_max)
+                                 min_splitable_entry_num=min_splitable_entry_num, min_rxns_to_spawn=min_rxns_to_spawn,extension_iter_max=extension_iter_max, extension_iter_item_cap=extension_iter_item_cap)
         else:
             def rxnkey(rxn):
                 c = 0