use a faster json library in lca/lca_db.py

[ctb]

perhaps one reason LCA databases are so slow is that we're using a slow JSON parser - the stdlib one!

ref #948 #909 and oddly #737

[luizirber]

perhaps one reason LCA databases are so slow is that we're using a slow JSON parser - the stdlib one!

orjson and rapidjson seem to be good drop-in replacements. hyperjson uses Rust underneath. rapidjson is in conda-forge already.
https://pythonspeed.com/articles/faster-json-library/

Or we could implement saving/loading in Rust... =]

[ctb]

I suspect that #1808 is going to be the best solution here - simply move to sqlite :)