bit/varbit stuff

I'm thinking of diving into the bit/varbit stuff, mainly because it's
practically the last area left on the list of functions not converted
to new fmgr style.

In a quick look at the code, I wonder why we are bothering with separate
bit and varbit functions/operators when they all do the same things.
Why not have one set of functions and mark the two types
binary-equivalent?

I notice also that we're missing a length-coercion function for bit.
I will add that, also the table entries needed for indexing support.

Adriaan, could I get your last set of changes (the patch submitted 7/21)
in the form of diffs from what you'd submitted before? It's not easy to
tell exactly what you changed --- diffing against what's in CVS isn't
very helpful because of the intervening pgindent reformat. Also, do you
object to sticking the standard Postgres copyright notices on the files?

regards, tom lane