reduce monorepo footprint

[jondot]

In order to make build folder take less space, and to ease up on rust-analyzer, we'd like to

• Move loco-cli to its own independent repo (this was part of the plan until it stabilizes)
• Move or remove macros, this was an experimental project that did not take off
• Move loco-extras to its own repo, make sure that we still have interconnectedness between loco and loco-extras in terms of breakages
• refactor + simplify xtask bump version process xtask: new bump versions #893
• add a 'publish' step for subcrates, then main crate via xtask

[joshka]

I really like that you have a mono-repo - it's a plus when you can just make changes and have them coherent with the rest of the repo. This is fairly small project all things considered - larger by at least two orders of magnitude git repos are still good for monorepos IMHO.

What's the problem with r-a that you're hitting?

[jondot]

@joshka you're right, better to keep a monorepo. right now i'm hitting a couple of limits:

• target/ folder size can reach 10-40GB on occasions
• rust-analyzer becomes sluggish as we're growing the list of members in the workspace

nevertheless, I've just added a subcrate: loco-gen because I had no other choice. For this instance, loco-gen will be the home of all things generators, including -- and more importantly -- generator testing, and generator testing was so heavy and filled with side-effects I had to isolate it in the form of a separate crate.

so looks like we're going to grow the amount of crates in the workspace any way, but then we'll remove some unneeded ones.

[jondot]

splitting remaining work into:

• xtask: new versioning + publishing workflow #913
• Unified lib+loco new + new starter generator #914