Merge back into hydra

[timkpaine]

Now that @omry has commit privileges again, hydra may be updated and become supported again. Since lerna only exists because facebookresearch let hydra become stale and unmaintained, we should consider merging our code upstream as it is a superset of functionality without much additional opinionation.

[omry]

Hi Tim.
Nice to meet you.
It will be some time before Hydra is fully back on its fit, but it will get there.
My initial focus is on OmegaConf, then Hydra.

I haven't taken a close look at Lerna so I don't have an opinion to offer on this yet.
Can you give me the high level view, what problems did solve, how etc?

[timkpaine]

I started by doing a straight 1-1 reimplementation in rust, replacing omegaconf with a rust variant and antlr with a rust parser. Then I took the top 10 or so issues from the hydra repo and resolved them, mostly to do with paths, defaults list, and list manipulation. All these "fixes" are in the README. Redoing it in rust wasn't strictly necessary but it's a healthy speed up and I didn't want to just hard fork without offering some type of compelling advantage. That being said I don't care too much about any of this work, I'd be happy to contribute all of this back.

[omry]

Gotcha.
I am not very familiar with Rust. adding it as a dependency instead of antlr is swapping one bad dependency for a potentially other bad one.
What is your take on portability, deployment (Windows, Linux, Mac X various CPU architectures)?
Hydra being pure Python means its very portable.