Add a persistent image cache

[CosmicHorrorDev]

This will eventually resolve #234

[CosmicHorrorDev]

Just some random tidbits:

• Useful looking crate for handling HTTP cache semantics (http-cache-semantics). Potentially this crate as a higher level one depending on how much async is a requirement there (http-cache)
• There's some interop required to get the right info out of ureq but that's not the worst
• The DB format is a bit more static with redb than it would be with some SQL DB that can be changed with migrations. I think we can have some consistent header with the internal version and TTL so that we can mix-n-match versions a bit as long as the table layout doesn't change

[CosmicHorrorDev]

thread 'image::cache::tests::layers' panicked at src/image/cache/tests.rs:306:66:
called `Result::unwrap()` on an `Err` value: Failed creating database

Caused by:
    Database already open. Cannot acquire lock.
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace


failures:
    image::cache::tests::layers

Guess who has two thumbs and gets to port things away from redb? 👍😞👍

[CosmicHorrorDev]

Still a little ways to go, but getting closer at least 🎉

running 1 test
test image::cache::tests::layers ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 52 filtered out; finished in 0.03s

[CosmicHorrorDev]

The test suite is a good representation of the current state of the PR

$ cargo t image::cache
...

running 17 tests
test image::cache::tests::global_ttl ... ignored, TODO: waiting for garbage collection
test image::cache::tests::lru ... ignored, TODO: waiting for garbage collection
test image::cache::tests::mutli_client_mash ... ignored, TODO: waiting for garbage collection
test image::cache::tests::remote_404_error ... ok
test image::cache::tests::private_cache ... ok
test image::cache::tests::etag_refresh_same ... ok
test image::cache::tests::selectively_stores ... ok
test image::cache::tests::remote_layers ... ok
test image::cache::tests::corrupt_db_entry ... ok
test image::cache::tests::etag_refresh_different ... ok
test image::cache::tests::local_invalidation ... ok
test image::cache::tests::invalid_img ... ok
test image::cache::tests::stats ... ok
test image::cache::tests::local_svg_layers ... ok
test image::cache::tests::local_layers ... ok
test image::cache::tests::past_max_age_refetch ... ok
test image::cache::tests::remote_svg_layers ... ok

test result: ok. 14 passed; 0 failed; 3 ignored; 0 measured; 54 filtered out; finished in 0.27s

The only thing left to implement is garbage collection (including probably updating entry's m-times) and wiring up the cache to actually be used by the application

Potential future work

• Integrating a lot more metrics
  • Counters for image loads (from source / l1 cached / l2 cached / errors)
  • Timings for image loads
• Switching DB connections to use a pool to manage re-using connections
• Make the stats output a lot prettier
• Switch the generation to a content hash