fix: Use built-in async for mqtt

[Nerivec]

• Use built-in async version of mqtt functions, should fix some promise awaiting weirdness (also allows a nice clean up of tests...)
• Add logging on publish error
• Fix typing
• Add more coverage
• Add handler for disconnect event (when using version: 5)
• Set maximum packet size in CONNECT packet (from settings)

Also tested manually on mosquitto and nanomq on top of jest tests...

@Koenkk Some notes:

• v5 features need a closer look from someone with a better knowledge of the MQTT spec (could be in follow-up PR though)
  • DISCONNECT seems to not really be respected by brokers (can't get it to trigger from nanomq when exceeding maximum packet size, it just drops the packet)?
  • might need more handling for disconnect event... might not reconnect?

Ref: https://www.emqx.com/en/blog/mqtt5-new-features-reason-code-and-ack

Re #24625

CC: @corporategoth @mundschenk-at

Docs Koenkk/zigbee2mqtt.io#3228

[corporategoth]

I should note that the default max packet size on EMQX is 1mb. So probably want to lower the default size we send to 1mb.
Also, might not be a bad idea to make that a configurable option in the MQTT settings.

[mundschenk-at]

I think the side effects would mainly be memory-related, i.e. buffer size both on the broker and client sides.

[Nerivec]

With the configurable setting, and the logging on disconnect with reason, it should be easy enough to track down and fix improper configurations. Allowing to leave the default at 1MB (which already amounts to roughly 3.5KB per device * 300 devices for /bridge/devices... should cover most networks with ease).

FYI, broker defaults seem to be:

• mosquitto: no limit (i.e. 256MB)
• emqx: 1MB
• nanomq: 2MB

[Koenkk]

This method logs logger.info('Connected to MQTT server'); while it may not be the case yet, I'm also not sure about the subscribe before it gets connected?