Add func call so we don't instruction-step into the builtin_trap

The way this test is structured right now, I set a breakpoint on
the instruction before the __builtin_trap.  It hits the breakpoint,
disables the breakpoint, and instruction steps.  This hits the
builtin_trap instruction which debugserver (on arm64) now advances
to the next instruction and reports that address to lldb.  lldb
doesn't recognize this as a proper response to the instruction
step and continues executing until the next trap, and the test fails.