Fix issue with failing CI tests (too early call to get_version of api.hive.blog instead of local testnet node)

Example job with such a fail: https://gitlab.syncad.com/hive/clive/-/jobs/1003688

Because of api.hive.blog updated to newer version than we're actually using in clive - we had tests failing because of too early call to node get_version.

It looked like this:

1. Initialize World
2. world.enter
3. node.setup
4. call do get_version (because of __sync_node_version) <--- fail here
5. Actual setting node address to the testnet one via node.set_address in conftest.py

Now, node.setup method is removed so get_version call via __sync_node_version is postponed until actually required (when getting network_type)