Crash the system server if it cannot bind to the network stack, or if
the network stack crashes. This is better than keeping the system
running in a bad state where all networking would be broken.
Before the network stack was moved to a different process, the whole
system would go down when it encountered a crash.
Test: added crash in NetworkStack, verified system_server restart
Bug: 130028724
Merged-In: I64f366fc69b52f27d65461be50d81edc199bcb62
Merged-In: I034a62ec44ba1da32fd47c842a2bdf86f6ccf542
(cherry picked from commit d44fc02248)
Change-Id: I989703c9dda62bd58df413ca8949f9a25187e062
9164fabba4