Divan extra benchmarks

[FilipAndersson245]

Description

This PR builds on #12000 and adds a few more benchmarks.
one to benchmark how long it takes to eval std, and a few different pipeline commands.

User-Facing Changes

Tests + Formatting

After Submitting

[kubouch]

I think it's a good idea to add more benchmarks. Some of the pipeline benchmarks seem quite random, though. For example, the ls depends on the directory contents, which might change. Ideally, the benchmarks should be written with some specific goal in mind. I like the stdlib and config loading and the serialization benchmarks, that's very useful.

[FilipAndersson245]

I think it's a good idea to add more benchmarks. Some of the pipeline benchmarks seem quite random, though. For example, the ls depends on the directory contents, which might change. Ideally, the benchmarks should be written with some specific goal in mind. I like the stdlib and config loading and the serialization benchmarks, that's very useful.

Yee the ls one was mostly as an initial test I had, I still think its interesting to track performance of simple IO command as they are so commonly used, I think they should be benchmarked to ensure they are snappy. I think the test of a simple for loop is quite straightforward, but the other one I more or less copied from this script inside nu_scripts.

[sholderbach]

After fixing conflicts this seems pretty straightforward to get into a shape so we can land those additions and iterate on our benchmarks, right?

[FilipAndersson245]

After fixing conflicts this seems pretty straightforward to get into a shape so we can land those additions and iterate on our benchmarks, right?

Yes, I have just been busy this last week will try to fix it today. :)

[FilipAndersson245]

This PR is ready for review.

[sholderbach]

I think we don't have to be perfect here yet and can observe and improve our benchmarks so we really draw some conclusions about future design decisions.

Thanks for moving us in the right direction @FilipAndersson245

[sholderbach]

Does using ignore here have a relevant impact?

[FilipAndersson245]

I am not sure, the problem is to get rid of the output that is displayed so it doesn't clutter the screen, if there is another way to do it I'm all for it.

[sholderbach]

shrugs at the limited number of core's on github workers

[FilipAndersson245]

Do you think this should be removed or changed?, even if cores are low just allowing multiple commands to be run parallel allow some level of parallelism even if its run on 1 thread machine. I initially added this test as I saw 1thread par-each outperform the standard each for low sample counts.