Functional improvements needed to VCH endpoint restart policy

[kjplatz]

Test setup:

1. Login to VCH appliance
2. Remove/rename /sbin/port-layer-server
3. Kill the port-layer-server

(Note: this works with the docker-engine-server and the vicadmin processes as well)
Expected behavior:
System attempts to restart a finite number of times, and then reports an error, leaving the VCH up and running in order to download logs

Actual behavior:
VCH immediately reboots.

[mdubya66]

Moving to Icebox per 9/20 Triage

[stuclem]

Already in the release notes, so removing the kind/note tag.

[hickeng]

Immediately rebooting is intentional to attempt to restore function - this should look into:

• marking the VCH as having had a failure of some kind
• suppressing subsequent restarts if the first restart was not beneficial
• suppressing restart at all if deployed with debug