Adding new environment files for stability

[sit23]

We have recently run into problems with the latest version of the openmpi package that is installed via conda ('openmpi=5.0.1'), which caused various glibc errors:

2024-01-24 12:28:12,404 - isca - INFO - /emmy-noether/home/sit204/mambaforge/envs/isca_env_jan_24/bin/../lib/gcc/x86_64-conda-linux-gnu/12.3.0/../../../../x86_64-conda-linux-gnu/bin/ld: /emmy-noether/home/sit204/mambaforge/envs/isca_env_jan_24/lib/libmpi.so: undefined reference to memcpy@GLIBC_2.14'`

2024-01-24 12:28:12,406 - isca - INFO - /emmy-noether/home/sit204/mambaforge/envs/isca_env_jan_24/bin/../lib/gcc/x86_64-conda-linux-gnu/12.3.0/../../../../x86_64-conda-linux-gnu/bin/ld: /emmy-noether/home/sit204/mambaforge/envs/isca_env_jan_24/lib/./libpmix.so.2: undefined reference to clock_gettime@GLIBC_2.17'`

To combat this, I've updated the main environment file to only use openmpi version 4, as this appears not to have this problem.

As a second item in the p/r, I've added a frozen version of the environment file for python 3.12, as this seems common amoungst other python packages. It is a configuration we have tested, and it seems to work stably.

The final change here is a move away from python 3.9 to 3.12, and a corresponding change in the test yml file (which is now set to use the frozen version of the environment file).

[sit23]

Using the frozen environment file for the ci test comes with the advantage that the ci doesn't have to spend ages solving the environment.

[dennissergeev]

Everything looks good to me from the Python perspective but I will defer a more strict review to a more active Isca user 😄

[ntlewis]

Happy for this to go in. With time, it would be better to patch issues introduced by changes in python package versions within the Isca code, but currently this isn't feasible given the time commitments of the Isca developers.