Skip Performance Improvement

[M-Schwab]

Adds an implementation of the Iterator method nth() to improve performance of skip, as well as transposes dir_values for better cache locality for higher dimensions.

From my initial tests, performance of skip is around a order of magnitude better when using dims==1000, though there is a fair bit of run/run variance.

Benchmarks

Skip benchmarks with both changes applied:

test bench_skip_f32  ... bench:     119,895 ns/iter (+/- 679)
test bench_skip_f64  ... bench:     204,413 ns/iter (+/- 1,197)
test bench_skip_i128 ... bench:     647,025 ns/iter (+/- 2,918)
test bench_skip_i16  ... bench:      58,755 ns/iter (+/- 146)
test bench_skip_i32  ... bench:     110,863 ns/iter (+/- 606)
test bench_skip_i64  ... bench:     190,389 ns/iter (+/- 1,264)
test bench_skip_i8   ... bench:       5,415 ns/iter (+/- 39)
test bench_skip_u128 ... bench:     666,928 ns/iter (+/- 5,908)
test bench_skip_u16  ... bench:      64,398 ns/iter (+/- 211)
test bench_skip_u32  ... bench:     104,274 ns/iter (+/- 431)
test bench_skip_u64  ... bench:     192,021 ns/iter (+/- 1,606)
test bench_skip_u8   ... bench:       5,407 ns/iter (+/- 30)

Skip benchmarks with only nth() implementation:

test bench_skip_f32  ... bench:     971,085 ns/iter (+/- 1,977)
test bench_skip_f64  ... bench:   1,027,240 ns/iter (+/- 10,761)
test bench_skip_i128 ... bench:   1,565,479 ns/iter (+/- 7,819)
test bench_skip_i16  ... bench:     944,982 ns/iter (+/- 2,097)
test bench_skip_i32  ... bench:     949,815 ns/iter (+/- 5,469)
test bench_skip_i64  ... bench:   1,017,815 ns/iter (+/- 10,016)
test bench_skip_i8   ... bench:     222,445 ns/iter (+/- 793)
test bench_skip_u128 ... bench:   1,544,590 ns/iter (+/- 5,082)
test bench_skip_u16  ... bench:     943,580 ns/iter (+/- 2,294)
test bench_skip_u32  ... bench:     953,715 ns/iter (+/- 5,762)
test bench_skip_u64  ... bench:   1,020,136 ns/iter (+/- 7,373)
test bench_skip_u8   ... bench:     220,886 ns/iter (+/- 938)

Skip benchmarks with no changes:

test bench_skip_f32  ... bench:   1,121,124 ns/iter (+/- 16,917)
test bench_skip_f64  ... bench:   1,307,425 ns/iter (+/- 9,478)
test bench_skip_i128 ... bench:   2,023,980 ns/iter (+/- 18,240)
test bench_skip_i16  ... bench:   1,000,306 ns/iter (+/- 14,314)
test bench_skip_i32  ... bench:   1,057,683 ns/iter (+/- 17,767)
test bench_skip_i64  ... bench:   1,164,238 ns/iter (+/- 14,959)
test bench_skip_i8   ... bench:     228,253 ns/iter (+/- 3,651)
test bench_skip_u128 ... bench:   1,797,844 ns/iter (+/- 8,894)
test bench_skip_u16  ... bench:     993,490 ns/iter (+/- 15,821)
test bench_skip_u32  ... bench:   1,042,425 ns/iter (+/- 15,699)
test bench_skip_u64  ... bench:   1,162,531 ns/iter (+/- 15,850)
test bench_skip_u8   ... bench:     231,441 ns/iter (+/- 3,570)