-fno-automatic option (PR 20592)

[Paul Brook]

> Where flag_automatic is the new flag corresponding to the option.
> Then, if it was that simple, it would have been done earlier, but even
> thinking harder I can't find what's wrong with this simple patch. Can
> someone help me? Is this the right thing, or did I miss an obvious
> point?

How does this patch interact with ALLOCATABLE variables. Are they deallocated 
at the end of the procedure or not? This needs testing and documenting.

Paul