implement advance_(back_)_by on more iterators

[the8472]

Add more efficient, non-default implementations for feature(iter_advance_by) (#77404) on more iterators and adapters.

This PR only contains implementations where skipping over items doesn't elide any observable side-effects such as user-provided closures or clone() functions. I'll put those in a separate PR.

[rust-highfive]

r? @joshtriplett

(rust-highfive has picked a reviewer for you, use r? to override)

[the8472]

@timvermeulen I changed the API contract of a bit. Do they look ok to you?

[SkiFire13]

I've left a couple of comments, let me know what you think about them

[the8472]

Thanks for the reviews @pickfire @SkiFire13 I have closed the conversation for the issues that I think should be addressed now. I have completely rewritten Skip::advance_by so you might want to give it another look.

The fixes are in separate commits for easier review.

[the8472]

I ran some of the core benchmarks, decent improvements for the .cycle().skip() case.

# RUSTFLAGS="-Ccodegen-units=1 -Copt-level=3" taskset -c 5 schedtool -B -e ./x.py bench --stage 0 library/core/ --test-args "cycle"

 iter::bench_cycle_skip_take_ref_sum          1,837,376         1,110,352              -727,024  -39.57%   x 1.65 
 iter::bench_cycle_skip_take_sum              1,297,196         525,064                -772,132  -59.52%   x 2.47 
 iter::bench_cycle_take_ref_sum               1,096,240         1,093,491                -2,749   -0.25%   x 1.00 
 iter::bench_cycle_take_skip_ref_sum          588,162           717,214                 129,052   21.94%   x 0.82 
 iter::bench_cycle_take_skip_sum              550,002           474,185                 -75,817  -13.78%   x 1.16 
 iter::bench_cycle_take_sum                   524,600           448,794                 -75,806  -14.45%   x 1.17 
 iter::bench_skip_cycle_skip_zip_add_ref_sum  5,336,791         3,761,926            -1,574,865  -29.51%   x 1.42 
 iter::bench_skip_cycle_skip_zip_add_sum      4,075,445         4,038,599               -36,846   -0.90%   x 1.01 

# RUSTFLAGS="-Ccodegen-units=1 -Copt-level=2 -Ctarget-cpu=native" taskset -c 5 schedtool -B -e ./x.py bench --stage 0 library/core/ --test-args "cycle"

 iter::bench_cycle_skip_take_ref_sum          1,396,256         1,087,191              -309,065  -22.14%   x 1.28 
 iter::bench_cycle_skip_take_sum              1,260,624         474,665                -785,959  -62.35%   x 2.66 
 iter::bench_cycle_take_ref_sum               1,095,467         1,094,409                -1,058   -0.10%   x 1.00 
 iter::bench_cycle_take_skip_ref_sum          1,058,240         1,008,308               -49,932   -4.72%   x 1.05 
 iter::bench_cycle_take_skip_sum              496,065           448,904                 -47,161   -9.51%   x 1.11 
 iter::bench_cycle_take_sum                   469,290           466,544                  -2,746   -0.59%   x 1.01 
 iter::bench_skip_cycle_skip_zip_add_ref_sum  3,453,381         3,737,632               284,251    8.23%   x 0.92 
 iter::bench_skip_cycle_skip_zip_add_sum      3,179,488         3,472,014               292,526    9.20%   x 0.92 

The zip case is interesting

    (0..100000).skip(100).cycle().skip(100)
      .zip((0..100000).cycle().skip(10))
      .map(|(a,b)| a+b)
      .skip(100000)
      .take(1000000)
      .sum()

The outer skip will call nth but that can't benefit from advance_by because as long as we have to preserve side-effects map has to use the default implementation for that. But the zipped iterators contain Skip too, which do benefit from advance_by. But that seems very opt-level-dependent.

[bors]

☔ The latest upstream changes (presumably #87768) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #87375) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #89158) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #89386) made this pull request unmergeable. Please resolve the merge conflicts.

[joshtriplett]

This looks reasonable to me.

@bors r+

[bors]

📌 Commit ffd7ade has been approved by joshtriplett