CLISP gets crazy results on certain rational args to `expt'

Here's an example

Reported in more detail on 2021-Mar-5 to https://sourceforge.net/p/clisp/bugs/744/

GNU CLISP 2.49.60+ (2017-06-25) (built on lgw01-amd64-012.buildd [127.0.1.1])

(setq rat 460536752677/8796093022208)

(expt 2 rat) => 1.0 (!)

(expt 3 rat) => 1.0