multiplies by powers of two

[Stephen Weeks]

> I claim that changing the basis library to use non-overflow-detection ops is
> bad.  It will, of course, speed things up, but the code should be written to
> be readable and the compiler should be figuring this out.

Too much of it is too hard.  Someday we will get some of it, but I think it is
the right balance for now to let a user run with -detect-overflow true on their
code but -detect-overflow false on the basis library code, which is presumably
much more thoroughly tested.