[BUG] new emit_*_scope funcs

[Jan Hubicka]

> 
> I guess you didn't test these at all.  They don't set
> set the scope on the last insn.  This means for emitting
> a single insn, no scope settings occur at all.

I did tested these on the testcases failing previously and it did work.
Thinking about it, all places where I do use it (ifcvt, instruction
splitting and peepholing) usually involves multiple instructions.

Thanks for noticing that!
> 
> I discovered this while repatching my sequence killing
> patches into the mainline... I can fix it for you.
Thanks!
If it is problem, I can do it today as well.

Honza