RFA: Changes to allow extensions to operator set

[Andrew Cagney]

> 1. There was essentially duplicate code in prefixify_subexp and 
>    length_of_subexp.  I have factored out the duplication.   [In the
>    process, by the way, I happened to notice that BINOP_VAL was
>    defined only in length_of_subexp, and indeed that there appears to
>    be no code anywhere to handle its evaluation.]

If I understand it correctly, this refactoring (hey I get to use a 
buzword) making operator_length global and then calling it  from 
length_of_subexp is independant of the other stuff, so yes, it can go in.

Nice cleanup.

Andrew