Refactor models to reduce duplication

[tjkirch]

To be able to have models/lib.rs with shared code, we needed to stop symlinking the entire src directory and instead treat the variant-specific code as Rust modules, symlinking the correct one to variant::current so that lib.rs can re-export Settings. (We also have to symlink in variant/mod.rs so Rust knows it's part of the module hierarchy.)

Now, each variant defines a Settings structure using common sub-structures. Those sub-structures live in the main lib.rs, and the modeled types are now in a shared modeled_types module. No changes were made to the code, just its organization.

Documentation was updated to reflect these changes, including using cargo-readme to generate README.md now that there's a single source.

---

Reviewer's note: regarding the giant green/red code blocks - nothing in the implementation changed, it's just in a different place.

Testing done:

Confirmed that all workspaces/ unit tests pass, whether for aws-dev or aws-k8s.

Confirmed that local builds still create the symlink(s) correctly and get the right model.

Made an aws-k8s AMI, confirmed that the containerd config still has the CRI settings. It connected to my cluster and ran a pod OK.

Made an aws-dev AMI, confirmed that the containerd config was the smaller dev version. Ran a docker container OK.

[tjkirch]

To help visualize the new structure:

├── build.rs
├── Cargo.toml
├── README.md
├── README.tpl
└── src
    ├── aws-dev
    │   ├── defaults.toml
    │   └── mod.rs
    ├── aws-k8s
    │   ├── defaults.toml
    │   └── mod.rs
    ├── lib.rs
    ├── modeled_types
    │   ├── kubernetes.rs
    │   ├── mod.rs
    │   └── shared.rs
    ├── variant
    │   ├── current -> ../aws-dev
    │   └── mod.rs -> ../variant_mod.rs
    └── variant_mod.rs

[zmrow]

This is super cool

🎁