iorap: Handle binder failures by gracefully attempting to reconnect

Also hide down functionality behind setprops:
* Debug logging (log.tag.ForwardingService VERBOSE)
* Log.wtf instead of crashing (iorapd.forwarding_service.wtf_crash true)
* Don't connect to binder if iorapd is disabled (iorapd.enable true)

The primary benefit of this is that:
a) If iorapd acts funny (or crashes) for any reason, it doesn't take down system server.
b) If developer manually runs 'adb shell stop iorapd', it won't crash
system server, and it doesn't need a restart ('adb shell stop/start') to
re-establish the binder connection.

Reconnections are done by sleeping forever in a 1-second loop.

Bug: 72170747
Test: adb shell stop && adb shell start # and manually set adb shell props
Test: atest system/iorap
Change-Id: I789db8962dc8498b1028639274cd34a0310b84c1