Can't use pytorch 1.12.1 for Docker image

[GFJHogue]

Attempting to run the Docker image built with torch==1.12.1 gives the following error.

ImportError: libtorch_cpu.so: cannot enable executable stack as shared object requires: Invalid argument

This probably requires a Python upgrade, which in turn requires adding a build step or upgrade for biopython for which our current pinned version 1.78 doesn't offer wheels above Python 3.9.

For now we'll stick with 1.9.1 (which we have used long-term) for the Docker image.