```
$ cargo bench  --manifest-path=src/Cargo.toml  -p vasi-sync
     Running benches/scchannel.rs (src/target/release/deps/scchannel-ae8dc0b68617fce5)
ping pong               time:   [4.3334 µs 4.5088 µs 4.6953 µs]
Found 5 outliers among 100 measurements (5.00%)
  1 (1.00%) low severe
  2 (2.00%) low mild
  2 (2.00%) high mild

ping pong pinned        time:   [3.2115 µs 3.2179 µs 3.2259 µs]
Found 13 outliers among 100 measurements (13.00%)
  7 (7.00%) high mild
  6 (6.00%) high severe
```

```
$ cargo bench  --manifest-path=src/Cargo.toml  -p vasi-sync
     Running benches/scchannel.rs (src/target/release/deps/scchannel-ae8dc0b68617fce5)
ping pong               time:   [4.2460 µs 4.4094 µs 4.5736 µs]
                        change: [-4.5442% -0.5164% +3.7840%] (p = 0.81 > 0.05)
                        No change in performance detected.
Found 6 outliers among 100 measurements (6.00%)
  1 (1.00%) low severe
  4 (4.00%) low mild
  1 (1.00%) high mild

ping pong pinned        time:   [3.1927 µs 3.1973 µs 3.2016 µs]
                        change: [-1.0734% -0.8648% -0.6660%] (p = 0.00 < 0.05)
                        Change within noise threshold.
```

```
$ cargo bench  --manifest-path=src/Cargo.toml  -p vasi-sync
     Running benches/scchannel.rs (src/target/release/deps/scchannel-ae8dc0b68617fce5)
ping pong               time:   [4.1223 µs 4.2546 µs 4.3760 µs]
                        change: [-8.6267% -4.6473% -0.6051%] (p = 0.03 < 0.05)
                        Change within noise threshold.
Found 7 outliers among 100 measurements (7.00%)
  3 (3.00%) low severe
  4 (4.00%) low mild

ping pong pinned        time:   [3.1942 µs 3.1974 µs 3.2008 µs]
                        change: [+0.0989% +0.2742% +0.4571%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 7 outliers among 100 measurements (7.00%)
  5 (5.00%) high mild
  2 (2.00%) high severe

```

This mostly works, but:

* there's a loom failure I can't figure out in one test (which
shouldn't actually matter in shadow)
* we can't safely drop the contents of the channel (if any)
after closing it. This might be fixable, but generally there's a
potential race if we try to drop the contents when closing, since it
could be getting read in parallel.

It appears to hurt performance without pinning, and be a no-op with
pinning.

```
$ cargo bench  --manifest-path=src/Cargo.toml  -p vasi-sync
ping pong               time:   [4.8376 µs 4.8934 µs 4.9665 µs]
                        change: [+10.053% +13.523% +17.508%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 11 outliers among 100 measurements (11.00%)
  2 (2.00%) low mild
  1 (1.00%) high mild
  8 (8.00%) high severe

ping pong pinned        time:   [3.2015 µs 3.2051 µs 3.2094 µs]
                        change: [+0.5280% +0.6995% +0.8895%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 4 outliers among 100 measurements (4.00%)
  3 (3.00%) high mild
  1 (1.00%) high severe
```

This reverts commit 11050ef.

This doesn't appear to help performance, and appears to still have some
concurrency bugs according to loom and miri.

Keeping in the commit history for posterity.

This reverts commit df29378.

This complicates the design and doesn't appear to help performance.
Keeping in the commit history for posterity.

This reverts three commits:
* Revert "Update SelfContainedChannel benchmark"
* Revert "scchannel: remove Reading state"
* Revert "scchannel: remove Writing state"

It also adds a comment to SelfContainedMutex to consider revisiting.

The reverted change significantly complicates the design, especially if
we wanted to make safe APIs (e.g. by adding checked creation of `!Send`
Reader and Writer ends of the channel).

On my desktop machine it appears to have ~zero performance benefit.

On a larger NRL machine it had a little performance benefit, but mostly
only when pinning is disabled, which we expect to be rare in
large simulations since that already significantly hurts performance.

In case we want to consider revisiting, here is the performance benefit
on one of the large NRL machines:

```
HEAD is now at 2628adb Update SelfContainedChannel benchmark
   Compiling vasi-sync v0.1.0 (/home/rjansen/workspace/temp/shadow/src/lib/vasi-sync)
    Finished bench [optimized] target(s) in 1.58s
     Running unittests src/lib.rs (src/target/release/deps/vasi_sync-fa33eb5cd8eff33e)

running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

     Running benches/scchannel.rs (src/target/release/deps/scchannel-24425348ef654f1a)
ping pong               time:   [11.567 µs 11.955 µs 12.254 µs]
                        change: [-25.107% -22.604% -19.866%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 7 outliers among 100 measurements (7.00%)
  5 (5.00%) low severe
  1 (1.00%) low mild
  1 (1.00%) high mild

ping pong pinned        time:   [2.1631 µs 2.1662 µs 2.1693 µs]
                        change: [-3.7961% -3.5645% -3.3038%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 100 measurements (3.00%)
  2 (2.00%) high mild
  1 (1.00%) high severe
```