Fix: Nested Choose fails when linked to group output

this bug occurs when:
 - the choose node is connected directly to Group output
 - the Group is inside a Schema

Perhaps i should have opted to add a no-op node so that the choose node can go ahead and prepare. Who knows? maybe I willl choose that solution in the future

this bug fix seems to cause no problems in existing trees,
and it fixes the behaviour in old trees that did not work.

schema_solve.py

@@ -313,7 +313,7 @@ class SchemaSolver:
         from_node = self.schema_nodes[(*self.node.ui_signature, from_ui_node.bl_idname)]
         from collections import deque
         unprepared = deque(from_node.hierarchy_dependencies)
-        self.prepare_nodes(unprepared)
+        self.prepare_nodes(unprepared, frame_mantis_nodes)
         from .utilities import cap, wrap
         get_index = from_node.evaluate_input("Index", self.index) # get the most recent link
         # getting the link at self.index just saves the trouble of killing the old links
@@ -514,15 +514,22 @@ class SchemaSolver:
                 return False
         return True
 
-    def prepare_nodes(self, unprepared):
+    def prepare_nodes(self, unprepared, frame_nodes):
         # At this point, we've already run a pretty exhaustive preperation phase to prep the schema's dependencies
         # So we should not need to add any new dependencies unless there is a bug elsewhere.
         # and in fact, I could skip this in some cases, and should investigate if profiling reveals a slowdown here.
-        forbidden=set()
-        e = None
-        # forbid some nodes - they aren't necessary to solve the schema & cause problems.
+        from .misc_nodes import UtilityChoose, UtilityKDChoosePoint, UtilityKDChooseXForm
+        forbidden=set() # forbid some nodes - they aren't necessary to solve the schema & cause problems.
         while unprepared:
             nc = unprepared.pop()
+            if isinstance(nc, (UtilityChoose, UtilityKDChoosePoint, UtilityKDChooseXForm)): 
+                can_skip=True                 # NOTE: this bug can also be fixed by adding
+                for output in nc.outputs:     # a no-op node between the choose and the output.
+                    for l in output.links:    # try that instead if this causes problems.
+                        if l.to_node in frame_nodes.values():
+                            can_skip = False; break
+                    if can_skip == False: break # don't keep looking
+                if can_skip: continue # do nothing because its links are not ready.
             if nc.node_type == 'DUMMY_SCHEMA' and not self.test_is_sub_schema(nc):
                 forbidden.add(nc) # do NOT add this as a dependency.
             if nc in forbidden: continue # trying to resolve dependencies for.
@@ -670,7 +677,7 @@ class SchemaSolver:
             if node.node_type == 'DUMMY_SCHEMA' and (schema_len_in := node.inputs.get("Schema Length")):
                 for l in schema_len_in.links:
                     unprepared.append(l.from_node)
-            self.prepare_nodes(unprepared)
+            self.prepare_nodes(unprepared, frame_mantis_nodes)
 
         # We have to prepare the nodes leading to Array Input Get
         for ui_link in array_input_get_link:
@@ -700,7 +707,7 @@ class SchemaSolver:
                 unprepared.extend(from_node.hierarchy_dependencies)
             else:
                 raise RuntimeError(" 671 there has been an error parsing the tree. Please report this as a bug.")
-            self.prepare_nodes(unprepared) # prepare only the dependencies we need for this link
+            self.prepare_nodes(unprepared, frame_mantis_nodes) # prepare only the dependencies we need for this link
             # and handle the output by the specific type
             if isinstance(to_ui_node, (SchemaConstOutput, NodeGroupOutput)):
                 self.handle_link_to_constant_output(frame_mantis_nodes, self.index, ui_link,  to_ui_node)