fix: allow address reuseage on socket binding

[firstof9]

may or may not fix [Errno 98] Address in use errors.

[lyricnz]

I don't think this fixes the issue (at least on my PC): if I add await asyncio.sleep(1) to the middle of TpLinkESS.query() to make it slow enough to be repeatably concurrent, then run >1 at the same time:

    async def run_query(switch, op_code):
        logging.info("Querying %s for %s", switch["hostname"], op_code)
        switch[op_code] = await x.query(switch["mac"], op_code)
        logging.info("Querying DONE %s for %s", switch["hostname"], op_code)

    for switch in switches:
        logging.info("Querying %s", switch["hostname"])
        tasks.append(run_query(switch, "stats"))
        tasks.append(run_query(switch, "ports"))
        tasks.append(run_query(switch, "vlan"))

I still get the error:

2023-02-08 09:11:28,097 INFO     Querying TL-SG108E for ports
2023-02-08 09:11:28,098 INFO     Querying TL-SG108E for vlan
2023-02-08 09:11:28,098 INFO     Querying Simon-TL-SG105PE for stats
2023-02-08 09:11:28,098 INFO     Querying Simon-TL-SG105PE for ports
2023-02-08 09:11:28,098 INFO     Querying Simon-TL-SG105PE for vlan
2023-02-08 09:11:28,099 INFO     Querying Back-TL-SG105PE for stats
2023-02-08 09:11:28,099 INFO     Querying Back-TL-SG105PE for ports
2023-02-08 09:11:28,099 INFO     Querying Back-TL-SG105PE for vlan
Traceback (most recent call last):
  File "/Users/simonroberts/src/github/firstof9/tplink-ess-lib/tplink_ess_lib/network.py", line 56, in __init__
    self.r_socket.bind((Network.BROADCAST_ADDR, Network.UDP_RECEIVE_FROM_PORT))
OSError: [Errno 49] Can't assign requested address

During handling of the above exception, another exception occurred:
...
  File "/Users/simonroberts/src/github/firstof9/tplink-ess-lib/tplink_ess_lib/network.py", line 58, in __init__
    self.r_socket.bind(("", Network.UDP_RECEIVE_FROM_PORT))
OSError: [Errno 48] Address already in use

[lyricnz]

Also, fundamentally we can't do this concurrently (unless we use dynamic listen ports). Since the socket-reader consumes all packets (even if they're not intended for the current query/response) this will cause timeouts for other operations.

[firstof9]

Maybe a loop with a sleep until it can grab the port?

[lyricnz]

Brutal, but would work. Would still need a retry/timeout to prevent lockup if "strange things happen". We already have a context manager to make sure the socket is closed.