mesa_temp_cache behavior

Hello!

When experimenting with non-standard neutrino cooling, I have encountered a weird and non-reproducible behavior, with different results depending on the presence of `.mesa_temp_cache` or not. This even if the `.mesa_temp_cache` is empty.

While digging further, I got confused on how this temporary cache actually works.

A possible brute-force fix could be `adding` `rm -r .mesa_temp_cache` to the `./clean` script, but probably best to first understand why that folder gets created and how it is used.