Add compilation steps for 10x faster inference

[do-me]

Add compilation steps for more than 10x faster inference. Benchmark on an M3 Max:

Without compilation

time cargo run -- encode "Hello world" minishlab/potion-base-8M

0.46s user 0.07s system 62% cpu 0.846 total

With compilation

time model2vec-rs encode "Hello world" "minishlab/potion-base-8M"

0.03s user 0.01s system 90% cpu 0.054 total`

[Pringled]

Hey @do-me, thanks for making this PR, good catch! Since the crate is published on crates.io, I think a nicer way to update the readme would be something like this:

cargo install model2vec-rs

model2vec-rs encode "Hello world" minishlab/potion-base-8M

This is a bit nicer than cloning the package and compiling it manually, let me know what you think.

[do-me]

I thought I'd keep it similar to the previous command to have sort of comparable results for the benchmark.
But yes great idea, these lines are useful if you just want to test it out really quickly. I can push some changes to the this PR later but feel free to modify of course!

Also, can you reproduce the 10x performance boost on your device?

[Pringled]

Nice! Sure thing, feel free to push the changes whenever you have a chance. I do see the same performance. The benchmarks I ran were already compiled though so the performance numbers do stay the same. Perhaps a nice way to structure the readme is to do an "either" in the quickstart of either adding the crate (to use it in a package), or installing the crate, to use it with the CLI. And then in the CLI section it can just be model2vec-rs encode "Hello world" minishlab/potion-base-8M etc instead of cargo run.

[do-me]

Completely restructured the readme similar to the Python repo. Also fixed some minor errors (quickstart reference was not clickable; some shell code snippets had Rust linting), aded a few things like citations, badges etc.
Hope this looks appropriate to you!

[Pringled]

@do-me nice, this looks great, thanks a lot for the updates 😄 ! Good to go, merging!