Truncate exit code to u8, even if the real code doesn't fit in u8.

[jakobnissen]

This prevents JuliaLauncher from panicking if a user exits with an error code which doesn't fit in u8.
This is already the behaviour of Julia on Linux, where ccall(:jl_exit) can be called with Int32(-1), yet returns error code 255.

Closes #737

[StefanKarpinski]

Can we please change things around so that the Julia process is the parent and the updater is the child? The. This would be unnecessary as would all the logic for handling the termination of the Julia child process.

[davidanthoff]

I have an alternative fix for the underlying problem in #750. @jakobnissen what do you think of that?

@StefanKarpinski I don't understand your suggestion :) The updater is not in the picture here at all, and is never the parent process. This here is the julialauncher, which is what starts when you type in julia. It is in charge of a) figuring out which Julia version to start, and b) will periodically launch the updater as a child process in the background.

[StefanKarpinski]

#552:

I think we should change julialauncher to use execve() (or its moral equivalent, wherever possible) instead of fork() when launching the julia sub-process. Because this does not launch a child process but instead causes the current process to be replaced by the new process

This would mean, among other benefits listed there, that julialauncher doesn't need to propagate the exit code because it becomes the Julia process which then terminates as it normally would. Having the real julia process be the child of the julialauncher process is just less good in basically every way.

[davidanthoff]

Yeah, I'm just quite nervous about the interaction of execve and Rust packages, there seems to be a lot of potential for pain. And on Windows that won't ever work, and I'm also not keen to have such a massive difference there...

My sense right now is that the only way to really have a one-process solution that is architecturally similar on the various platforms would be some variant of #327. Which also seems anything but trivial...