Split into crates

[Alex-Fischman]

This PR splits the project up into multiple crates, which in theory will improve compile times. While this is a large change, most of the changes that I had to make were mechanical, except for the following:

• I made all mods in all lib.rs and mod.rs public. This causes lots of missing documentation warnings. I did this because a) the code does actually need to be documented, and b) I believe that all modules in a crate should be public, and the module should decide what is exposed. While the second point is debatable, I don't think the first point is, so I don't think that this change should be reverted until we find a way to get the clippy lints without it.
• I removed all of the existing instances of pub use in mod/lib.rs. Since all modules are now public, this just creates an extra path to the same code, so it's unnecessary.
• I added some pub uses to the lib.rs in solver, so that clients of solver would not have to also depend on smtlib.
• We split the code up into crates as follows: fly and smtlib define common functionality among other crates; bounded, inference, solver, and verify all implement the functionality that the user cares about; and temporal-verifier defines the command line interface. While most files are in the crate that corresponds to the folder that they were in before this commit, a few files (such as timing.rs and concurrent.rs) got moved due to being in the wrong place in the dependency ordering.
• Finally, there was one more folder, term that we weren't quite sure what to do with, as it defines common functionality but also depends on fly. For now, we've left it as a submodule of the fly crate.

[Alex-Fischman]

I've left unused lines commented out in the root Cargo.toml. Someone who added those lines should look at them and determine if we still need them or not.

[wilcoxjay]

This is great, thanks so much for taking this on @Alex-Fischman!

As a side note, I did some further investigation into build times, and a significant portion (~50% by my estimate) of the end-to-end build time is spent recompiling the qalpha branch of App::exec. I believe this is due to the heavy use of generic instantiation in that branch. In the future it would be great to move that code into the qalpha file so that it does not need to be recompiled every time we build the app.

[Alex-Fischman]

I'm currently working on getting CI to pass by adding documentation, but I haven't used the code that I'm trying to document, so someone should check my work when I'm finished.

[Alex-Fischman]

Whoops, CI could've passed without getting docs done. There's some issue with tests in proc.rs. I'll look into it tomorrow.

[tchajed]

Could you stop adding any documentation in this PR, and turn off #[warn(missing_docs)] for now? It makes the PR much harder to review.

[Alex-Fischman]

Could you stop adding any documentation in this PR, and turn off #[warn(missing_docs)] for now? It makes the PR much harder to review.

Sure, I'm just going to commit my documentation for MARCO, since I wrote the file. Then I'll comment out the #[warn] and open an issue to add it back in.