test: Avoid redundant stop and error spam on startup failure

Trying to immediately shut down a node after a startup failure without
waiting for the RPC to be fully up will in most cases just fail and lead
to an RPC error.

Also, it is confusing to sidestep the existing fallback to kill any
leftover nodes on a test failure.

So just rely on the fallback.

Author: MarcoFalke

test/functional/test_framework/test_framework.py

@@ -567,15 +567,10 @@ def start_nodes(self, extra_args=None, *args, **kwargs):
         if extra_args is None:
             extra_args = [None] * self.num_nodes
         assert_equal(len(extra_args), self.num_nodes)
-        try:
-            for i, node in enumerate(self.nodes):
-                node.start(extra_args[i], *args, **kwargs)
-            for node in self.nodes:
-                node.wait_for_rpc_connection()
-        except Exception:
-            # If one node failed to start, stop the others
-            self.stop_nodes()
-            raise
+        for i, node in enumerate(self.nodes):
+            node.start(extra_args[i], *args, **kwargs)
+        for node in self.nodes:
+            node.wait_for_rpc_connection()
 
         if self.options.coveragedir is not None:
             for node in self.nodes: