default to coredumping instead of just in time debugging

JIT debugging has a number of challenges, chief among them is that we
have to close all FDs to facilitate auto restarts, but doing so opens
other threads up to stumble over those then-closed FDs and crash while
crash handling. Instead of JIT we do a re-raise after having written the
necessary metadata and having closed the FDs. This still has a bit of
racing potential but it's in the scale of microseconds rather than
actual seconds (drkonqi has a builtin delay on top of the fork-exec
overhead)

This has already been used in KDE neon unstable for years without
negative reports so should be save to use as default.

Re-raising is only applicable on Linux where we'll forward the crash
into potential handler processes (e.g. coredumpd, abrt, apport ...) in
the coredumpd case we also take care of presenting a UI via drkonqi.

For systems without handler support we continue to attempt JIT by
default. Though honestly that needs revisiting too because arguable the
only case where JIT is actually required is on freebsd for lack of a
better handling alternative that I know of. All other platforms offer
some form of crash handling already so our JIT stuff is mostly just
getting in the way.