Deadlock (busy loop) on panic/SIGSEGV case in Go API user code

Hello.

With the recent SA_ONSTACK fix I get deadlock in this minimal Go example code:

```
package main
/*
#cgo pkg-config: GraphicsMagick
#include <magick/api.h>
*/
import "C"
import "fmt"
type A struct { b *B }
type B struct { c int }
func main() {
	C.InitializeMagick(nil)
	a := A{nil}
	fmt.Println(a.b.c)
}
```

Without SA_ONSTACK fix I get this output which is not normal Go panic too:
https://gist.github.com/f0f4a1b0e65021050f020b3d88b06aaa

Normal panic should look like this (comment "C.InitializeMagick(nil)" line):

```
$ go run gm.go
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x483937]

goroutine 1 [running]:
main.main()
	/home/kagami/gm.go:13 +0x27
exit status 2
```

Probably it's caused by violating some rules described at:
https://golang.org/pkg/os/signal/#hdr-Go_programs_that_use_cgo_or_SWIG

E.g. SA_RESTART and signal mask bits. Maybe it would be better to not redefine
"oldact"? I.e. make first sigaction call with act=NULL and don't re-set it in
case if it's non-empty.

It would be even better to not set any handlers for API users at all, unless
requested, but as you said before it might cause ABI change...

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot