Increase startup reset wait to 2 seconds (#714)

When connecting over a TCP serial socket (e.g. the silabs-multiprotocol
add-on), the NCP reset frame can take ~1.5s to arrive after the remote
reuses the socket from a previous instance. The 1s wait causes a
spurious retry on every startup; bumping to 2s lets the reset land
within the first attempt.

Extend `EmberStatus` -> `sl_Status` mapping (#707)

* Fix final few uses of EmberStatus

* Map out as many EmberStatus -> sl_Status mappings as possible

Revert `TABLE_ENTRY_ERASED` change

Reduce log level of NV3 interface warning (#702)

* Reduce log level of NV3 interface warning

* Update test

Move `ezspVersion` key to correct place for Z2M (#701)

* Move `ezspVersion` key to correct place for Z2M

* Update tests

Implement zigpy TX power API (#698)

* Implement zigpy TX power API

* Fix typo in command name

* Bump zigpy

* Add tests

Reduce `RST` cancel prefix to 32 bytes (#697)

Reduce RST cancel prefix to 32 bytes

Move retrying out of `connect` and into `startup_reset` (#696)

* Move retrying out of `connect` and into `startup_reset`

* Constant tweaking

* Better debug logging for long CANCEL sequences

* Fix tests

Backup/restore routing table and allow specifying TX power during for…

…mation (#690)

* Restore the routing table on startup

* Account for new zigpy changes

* Read the route table size

* Rename `RouteRecordStatus` constants

* Bump zigpy to 0.85.0

* Add unit tests

* Re-add accidentally removed code

* Clean up

Send three resets when connecting, not just one (#689)

* Send three resets when connecting, not just one

* Fix unit test

Handle bad values in XNCP chip info parsing (#688)

* Fail gracefully when parsing invalid XNCP chip info response

* Add a test