GDB should skip SIGILL for wrfsbase

[davidchisnall]

Since the changes to OE to support emulating the wrfsbase instruction, we now get a SIGILL on every lthread switch. This causes gdb to stop. Ideally, the gdb scripts should check if the faulting instruction is wrfsbase and, if so, resume automatically (or, perhaps, turn off tracing on SIGILL entirely and instead automatically put a breakpoint on the uncommon paths in the OE signal handler).

[jxyang]

@anakrish I remember we had a patch for skipping SIGILL for cpuinfo. Was it merged?