Reject starting with multiple threads if --handle-signals=no

[vchuravy]

Julia GC safepoint is based on a read fault on a protected page. If you start Julia with multiple threads and pass --handle-signals=no the signal handler will cause a unhandled segmentation fault.

I don't know if we could install a "partial" segmentation fault handler that only handles safepoints,
but we should not let users shoot themselves into the foot. An easy solution is to document that multi-threading requires the signal-handlers and to not allow starting with multiple threads when --handle-signals=no.

This is of course only a partial fix, since foreign threads may become adopted nontheless.

As seen in the wild: https://discourse.julialang.org/t/floop-threading-and-juliacall-produce-segmentation-fault/100717/11

x-ref: JuliaPy/PythonCall.jl#219

[vtjnash]

Or add something like libjsig to our distribution? https://docs.oracle.com/javase/8/docs/technotes/guides/vm/signal-chaining.html

[vchuravy]

Yeah that would make cases like #48378 workable, and maybe better the MPI situation as well.