CI: Various CI fixes

• I merged a change without updating the author list. This has happened a few times and I'm not sure why. I suspect it is an issue with the headless checkout used in CI. We should look into how to make this more reliable in future.
• This change also seems to have introduced a bug that the CI missed before. Nothing is being returned when reading a frame file but not resampling.
• An upstream change in lalsimulation EoS tools has broken another test. I made the numeric test less stringent to avoid being dinged by this again.
• There were then many other breaking tests that seemed to stem from using new versions of numpy with old versions of numba/pymc. This will likely sort itself out upstream, but I manually pinned minimum versions of numba/pymc to help the conda version resolver.
• Debugging the dockerfiles was a massive pain so I decided to with to a conda environment file and installation using mamba. The new method runs ~5x faster than the old method. I think this is a natural step in the migration from pypi -> conda/mamba environment management.