Ticket 371: Fix broken definition of expt

R6RS says:

0. 0^z is 1.0 if z = 0.0, and 0.0 if (real-part z) is positive. For other cases in which the first argument is zero, either an exception is raised with condition type &implementation-restriction, or an unspecified number object is returned.

R5RS specifies behavior that is simply incorrect for z with negative real part: 0^z is 1 if z = 0 and 0 otherwise.

I tested the usual suite on (expt 0.0 0.0) (expt 0.0 2), and (expt 0.0 -2):

Conforming implementations:

Racket, Gauche, Gambit, Chicken (with and without the numbers egg), Bigloo, Kawa, Chibi, Chez, SCM, Ikarus/Vicare, Ypsilon, Mosh, NexJ, STklos, KSi, RScheme, XLisp, Rep, VX, SXM, Spark, Inlab return 1.0, 0.0, and +inf.0.

Guile, Larceny return 1.0, 0.0, and +nan.0.

SISC, IronScheme, Elk return 1.0, 0.0, and 0.0.

UMB returns 1.0, 0.0, and an exception.

Non-conforming implementations:

Scheme48, BDC return an exception, 0.0, and +inf.0.

Scsh returns -nan.0, 0.0, and an exception.

Scheme 7 returns 0.0, 0.0, and an exception.

Scheme 9 returns exact 0, 0.0, and +nan.0.

Elk returns exact 1, exact 0, and exact 0.

MIT returns an exception, 0.0, and another exception.

SigScheme, Shoe, TinyScheme, Dream, Schemik, Oaklisp, Owl Lisp don't support inexact expt.