fix: additional polling updates

[jrvb-rl]

• update sync polling to match the fixed async polling
• fixed up name of the async_retry_server_poll_until() function
• also avoid dup retry logic for long polling timeouts

The retry logic is particularly irksome. The HTTP wrappers in _base_client.py have their own retry logic which by default retries 40X responses. For long polls that use _post() this does the wrong thing, since it retries the exact previous request regardless of what the higher level polling loop might want. Skip retries in that case.