feat: Availability improvements

[Nerivec]

Important

A router failing to ping likely means a route in the network is lost, which can seriously affect the stability of the network if it happens on a regular basis (delays, missing state updates, other devices temporary unavailable, etc.).

Changes

• Add jitter to timeout to spread pinging a bit
  • Should prevent storming networks with successive pings
• Add backoff to timeout to prevent pinging devices as often when they previously failed
  • Backoff calc: *1.5 if device was previously available, *2 otherwise
    • resulting in: *1.5, *3, *6, *12... (with default timeout of 10min: 10, 15, 30, 60, 120)
  • Add ability to pause pinging after backoff reaches a certain value (resumed on next "last seen" update) - this is disabled by default (0) to not change the default Availability behavior too much
• Widen the definition of "active" to all power sources not unknown/battery to support devices possibly using 3-phase, DC & other more specific sources
• Cleanup various getters, use Map/Set instead of objects, and some more little fixes

Changes in Availability behavior

• Three new settings:
  • availability.active.max_jitter (default: 30000)
  • availability.active.backoff (default: true)
  • availability.active.pause_on_backoff_gt (default: 0)
• Pinging is now randomly offset from timeout by up to max_jitter
• Backoff is now enabled by default, successive failed pings will increase interval between tries automatically

Should alleviate some of the issues described in #23910 (and others)

TODO:

• fix: Availability changes in Koenkk/zigbee2mqtt#26811 zigbee2mqtt.io#3679
• confirm default settings