support caching all jitted functions

[jmoralez]

Adds the cache and nogil arguments to all jitted functions. These can be set with:

• NIXTLA_NUMBA_CACHE: if this variable is set all functions will be cached
• NIXTLA_NUMBA_RELEASE_GIL: if this variable is set all functions release the GIL.

Also adds a small notebook (nbs/docs/how-to-guides/numba_cache.ipynb) explaining the caching.

NOTE:
The nelder_mead function couldn't be cached because it received another function as input and I couldn't figure out a way to cache it like that, so it's currently repeated for ces, ets and theta (the only difference is the objective function, which is baked into the function).

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[AzulGarza]

awesome @jmoralez! LGTM😍

[VivCh14]

Hi, really great addition! Thank you for working on this.

I tried setting the environment variable 'NIXTLA_NUMBA_CACHE=1' or 'NIXTLA_NUMBA_CACHE=0' and they both tend to enable caching! To not cache I have to comment out the 'NIXTLA_NUMBA_CACHE' altogether. Any thoughts ?

[jmoralez]

The caching is enabled if the environment variable is set, so if you don't want to enable it you shouldn't set it or delete it if you've already set it elsewhere, e.g. del os.environ['NIXTLA_NUMBA_CACHE']

[VivCh14]

Thanks! makes sense :)