perf/fix: retain mut algorithm

[fereidani]

Closes #444

• fixes panic behavior to match rust standard library
• improves performance of retain algorithm by reducing memory copies

benchmark results shows about 20-30% performance improvement in cases that retain actually happens:

     Running benches/bench.rs (target/release/build/smallvec/5ddccd3aae9f1495/out/bench-5ddccd3aae9f1495)
bench_retain_mut_half   time:   [70.053 ns 70.416 ns 70.873 ns]
                        change: [-31.072% -29.765% -28.409%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 12 outliers among 100 measurements (12.00%)
  4 (4.00%) high mild
  8 (8.00%) high severe

bench_retain_mut_half_small
                        time:   [34.183 ns 34.212 ns 34.242 ns]
                        change: [+0.8531% +1.0178% +1.1919%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 5 outliers among 100 measurements (5.00%)
  4 (4.00%) high mild
  1 (1.00%) high severe

bench_retain_mut_all    time:   [27.246 ns 27.261 ns 27.277 ns]
                        change: [-5.8022% -3.9459% -2.4634%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 100 measurements (4.00%)
  1 (1.00%) low mild
  1 (1.00%) high mild
  2 (2.00%) high severe

bench_retain_mut_all_small
                        time:   [27.199 ns 27.233 ns 27.275 ns]
                        change: [+1.4440% +2.2589% +3.1175%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 22 outliers among 100 measurements (22.00%)
  4 (4.00%) high mild
  18 (18.00%) high severe

bench_retain_mut_none   time:   [27.183 ns 27.202 ns 27.221 ns]
                        change: [-0.6473% -0.2981% +0.0314%] (p = 0.09 > 0.05)
                        No change in performance detected.
Found 4 outliers among 100 measurements (4.00%)
  4 (4.00%) high mild

bench_retain_mut_none_small
                        time:   [27.180 ns 27.199 ns 27.217 ns]
                        change: [-4.1071% -2.2619% -0.7299%] (p = 0.00 < 0.05)
                        Change within noise threshold.

bench_retain_mut_vec_half
                        time:   [59.931 ns 60.049 ns 60.181 ns]
                        change: [-19.577% -19.106% -18.723%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 100 measurements (4.00%)
  4 (4.00%) high mild

bench_retain_mut_vec_half_small
                        time:   [28.589 ns 28.622 ns 28.662 ns]
                        change: [-3.1025% -2.1240% -1.2223%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
  4 (4.00%) high mild
  2 (2.00%) high severe

bench_retain_mut_vec_all
                        time:   [25.772 ns 25.800 ns 25.827 ns]
                        change: [+0.5852% +0.7517% +0.9227%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 8 outliers among 100 measurements (8.00%)
  1 (1.00%) low mild
  4 (4.00%) high mild
  3 (3.00%) high severe

bench_retain_mut_vec_all_small
                        time:   [25.933 ns 25.965 ns 26.005 ns]
                        change: [+0.8062% +0.9226% +1.0388%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 2 outliers among 100 measurements (2.00%)
  1 (1.00%) high mild
  1 (1.00%) high severe

bench_retain_mut_vec_none
                        time:   [25.625 ns 25.637 ns 25.651 ns]
                        change: [-1.8376% -1.7440% -1.6515%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high mild

bench_retain_mut_vec_none_small
                        time:   [25.897 ns 25.922 ns 25.948 ns]
                        change: [-0.2309% -0.0278% +0.2089%] (p = 0.82 > 0.05)
                        No change in performance detected.
Found 6 outliers among 100 measurements (6.00%)
  2 (2.00%) low mild
  1 (1.00%) high mild
  3 (3.00%) high severe

[alejandro-vaz]

yeah improvements are crazy

[alejandro-vaz]

the only downside is that the algorithm is significantly more complex but it is worthwhile

I'll re-review this tomorrow, these hours are not a good time to be reasoning about unsafe rust, but it looks good

[alejandro-vaz]

@TDecking what are your thoughts on this

[TDecking]

This PR is a port of a std::Vec::retain_mut change done several month earlier. It should be fine to include.

[fereidani]

Yes, I contributed this few months back to rust itself, It is already on stable Rust.
I just did some minor changes to match smallvec structure and fix some miri errors.