[R6RS] record-constructor suggestion

[dyb at cs.indiana.edu]

> > I suggest that we add an rtd argument to record-constructor and make the
> > record-constructor-descriptor argument optional.
>
> Any reason why you don't always accept a single argument and make a
> case distinction between an rtd and a constructor descriptor?

I was thinking of going further and eliminating the rtd argument from
make-record-constructor-descriptor, since I'm not sure the implementation
needs to know the rtd until record-constructor is called.  But that seemed
like too radical a change---especially with the other stuff we need to get
done---at this point, so I backed off.  I think I prefer to leave the door
open for this change in the future by requiring the rtd argument for
record-constructor, but I'm willing to go either way.

Kent