[RFC] x86_64: Simplify powl computation for small integral y [BZ #33411]

[Paul Zimmermann]

       Hi Siddhesh,

thank you for looking at this.

> This does not introduce any regressions other than the following:
> 
> testing long double (without inline functions)
> Failure: Test: pown_upward (-0x1p+8192, 3LL)
> Result:
>  is:         -inf  -inf
>  should be:  -1.18973149535723176502e+4932  -0xf.fffffffffffffff0p+16380
> Failure: Test: pown_upward (-0x1p-8192, -3LL)

since the result overflows towards -inf and rounding is upward, the result
should be the smallest regular number, not -inf. Thus it is indeed a regression.

Paul