Crashtracker: investigate how we can coexist with libsegfault

[danielsn]

          IMHO, this is the worst possible outcome.  I've actually observed this to occur in a situation where libSegFault.so is layered with crashtracking and the underlying process does some _something_ (I don't yet know).

It isn't clear to me how to prevent this, since it seems that all of the solutions come down to exiting the process or just avoiding responsibility by mutating the sigaction under some condition. I think the best idea I personally have is to use the equivalent of TSC to estimate the amount of off-signal work done between invocations of the handler, and if it's below some threshold then _exit() if the faulting IP matched the previous one.

Originally posted by @sanchda in #696 (comment)