Come up with a new version of `expt`

[jpellegrini]

@egallesio since you suggested that we had sqrt call expt, I have been looking into it, and I think it would make sense to rewrite expt completely, and indeed make sqrt call it. I'm working on it, and I opened this issue to let you know.

So far the first version of expt would work like the following (things may be changed later for more precision or efficiency, but this would be a good start):

static SCM my_expt_new(SCM x, SCM y) {
/*
      x         y     result
============================
----------------------------  y = 0:
                0     1
----------------------------  x = 0, 1:
       1              1
       0      < 0     error
       0      >=0     0
----------------------------  y = 1, 1.0:
                1     x
              1.0    inex(x)
----------------------------  x REAL:
] 0.0,+1.0[   +big   +0.0
]-1.0, 0.0[   +big   0.0 * parity(y)
   +real      +big   +inf.0
   -real      +big   +inf.0 * parity(y)
    real   +fixnum   expt_real_positivefixnum
    real      real   pow(x,y)
   +-0.0 +rational   0.0
   +-0.0 -rational   +inf.0
   +real +rational   pow(x,inexact(y)) or exp( log(x) * y )
   -real +rational   make-rect(0,  real_part( exp( log(x) * y ) ) )
    real   complex   exp( log(x) * y )
    real       < 0   1 / exp(x,-y)
---------------------------- x EXACT except 0, 1:
   exact      +big   error
   exact   +fixnum   expt_exact_positivefixnum
   exact   -fixnum   1 / exp(x,-y)
   exact  rational   TODO
   exact      real   make-rect(0,  real_part( exp( log(x) * y ) ) )
   exact   complex   exp( log(x) * y )
---------------------------- x COMPLEX:
 complex             exp( log(x) * y )

*/

[jpellegrini]

See #931