One more global.c speedup

[Brad Lucier]

>   > Perhaps now it's time to work on jump :-).
> I know what's going on in delete_from_jump_chain.  It's pretty obvious.  We
> For each jump, we have a list of other unconditional jumps which reach the
> same label.  Those lists are long because of the kind of code you are writing.

I don't quite understand.  The routine is one large switch statement, with
many labels; it doesn't use computed goto's.  There are really only two labels
that have a lot of jumps to them, one just before the switch statement for
internal jumps, and one at the end before the state cleanup and exit code;
all the other labels have only a "few" (I'd guess < 20) jumps to them.

Will this type of code cause this effect in jump.c?

Brad