...from 8046a87e.  At least sometimes in
sw_complex test (with various bundled Java extensions enabled) under load, this
code reports "Signal 11 during cups initialization called, ignoring cups" and
in-process JVM aborts in panic.  Looks like a legitimate SIGSEGV (to be
translated into java.lang.NullPointerException) from JVM code is erroneously
caught by the temporary lcl_signal_action in cupsmgr.cxx instead.  As there is
no non-cooperative way to have different signal handlers for different threads
(at least under POSIX), and
<https://issues.apache.org/ooo/show_bug.cgi?id=86306> "office crashes at startup
on Solaris Intel" suggests this signal-catching business is only there to work
around a completely broken machine, I think it is best to simply remove it
again.

Change-Id: I55b95a71d622f83c975989a4ffb1d95ef5737075