[tests] TestNode: separate add_node from start_node

Separates the act of creating a TestNode object from starting the node.
The test_framework now keeps track of its list of TestNodes, and test
writers can call start_node() and stop_node() without having to update
the self.nodes list.

test/functional/p2p-versionbits-warning.py

@@ -112,7 +112,7 @@ class VersionBitsWarningTest(BitcoinTestFramework):
         # Empty out the alert file
         with open(self.alert_filename, 'w', encoding='utf8') as _:
             pass
-        self.nodes = self.start_nodes(self.num_nodes, self.options.tmpdir, self.extra_args)
+        self.start_nodes()
 
         # Connecting one block should be enough to generate an error.
         self.nodes[0].generate(1)
@@ -123,7 +123,7 @@ class VersionBitsWarningTest(BitcoinTestFramework):
         self.test_versionbits_in_alert_file()
 
         # Test framework expects the node to still be running...
-        self.nodes = self.start_nodes(self.num_nodes, self.options.tmpdir, self.extra_args)
+        self.start_nodes()
 
 if __name__ == '__main__':
     VersionBitsWarningTest().main()