Add an optimized token router filter

[Jake-Shadle]

This PR introduces a HashedTokenRouter filter in parallel with the current TokenRouter that is significantly faster.

1. The current token router linearly searches through every token in every endpoint and does a full byte comparison. In contrast, the new token router instead hashes every token when a cluster's endpoints are updated and buckets the addresses by the hash, so that the lookup is a single operation.
2. Once 1 or more addresses are found by token in a locality, we abort the search, rather than keep searching, as this doesn't really make sense.

token_router                      fastest       │ slowest       │ median        │ mean          │ samples │ iters
╰─ token_router                                 │               │               │               │         │
   ├─ HashedTokenRouter                         │               │               │               │         │
   │  ├─ multi:2..128:duplicates  1.882 µs      │ 32.4 µs       │ 13.71 µs      │ 14.28 µs      │ 100     │ 100
   │  │                           56.27 TB/s    │ 3.27 TB/s     │ 7.724 TB/s    │ 7.417 TB/s    │         │
   │  ├─ multi:2..128:unique      1.332 µs      │ 41.31 µs      │ 13.8 µs       │ 15.21 µs      │ 100     │ 100
   │  │                           79.46 TB/s    │ 2.563 TB/s    │ 7.67 TB/s     │ 6.959 TB/s    │         │
   │  ├─ single:duplicates        791.7 ns      │ 20.56 µs      │ 6.602 µs      │ 6.8 µs        │ 100     │ 100
   │  │                           2.075 TB/s    │ 79.88 GB/s    │ 248.8 GB/s    │ 241.6 GB/s    │         │
   │  ╰─ single:unique            1.271 µs      │ 22.73 µs      │ 6.742 µs      │ 7.238 µs      │ 100     │ 100
   │                              1.292 TB/s    │ 72.29 GB/s    │ 243.7 GB/s    │ 227 GB/s      │         │
   ╰─ TokenRouter                               │               │               │               │         │
      ├─ multi:2..128:duplicates  31.77 ms      │ 65.87 ms      │ 61.48 ms      │ 59.32 ms      │ 100     │ 100
      │                           3.334 GB/s    │ 1.608 GB/s    │ 1.723 GB/s    │ 1.786 GB/s    │         │
      ├─ multi:2..128:unique      23.23 ms      │ 71.5 ms       │ 59.99 ms      │ 56.36 ms      │ 100     │ 100
      │                           4.557 GB/s    │ 1.481 GB/s    │ 1.765 GB/s    │ 1.878 GB/s    │         │
      ├─ single:duplicates        8.828 ms      │ 9.692 ms      │ 8.981 ms      │ 8.999 ms      │ 100     │ 100
      │                           186.1 MB/s    │ 169.5 MB/s    │ 182.9 MB/s    │ 182.5 MB/s    │         │
      ╰─ single:unique            8.824 ms      │ 9.566 ms      │ 9.027 ms      │ 9.051 ms      │ 100     │ 100
                                  186.2 MB/s    │ 171.8 MB/s    │ 182 MB/s      │ 181.5 MB/s    │         │

[quilkin-bot]

Build Succeeded 🥳

Build Id: 9d024cad-0864-4905-8941-86a584f78242

The following development images have been built, and will exist for the next 30 days:

• us-docker.pkg.dev/quilkin/ci/quilkin:0.9.0-dev-4f3ca26

To build this version:

git fetch git@github.com:googleforgames/quilkin.git pull/948/head:pr_948 && git checkout pr_948
cargo build