Standard deviation for time samples

[WhiteBlackGoose]

Hey, just started using your library today and I really like it. A lot better than the built-in thing.

One (imo vital) statistic that I found missing is the standard deviation, so started working on that one. It's not finished yet, but feel free to correct me on the way so I wouldn't waste time on something you don't want in the lib.

Alternatively to standard deviation we could do sample variance (in my opinion it's less readable, but I found it in the roadmap of this library).

Timer precision: 22 ns
bench1        fastest       │ slowest       │ median        │ mean          │ std dev │ samples │ iters
╰─ fibonacci                │               │               │               │         │         │
   ├─ 1       0.549 ns      │ 1.384 ns      │ 0.614 ns      │ 0.644 ns      │ 0.655 ns │ 100     │ 409600
   ├─ 2       1.593 ns      │ 1.937 ns      │ 1.66 ns       │ 1.691 ns      │ 1.693 ns │ 100     │ 204800
   ├─ 4       4.69 ns       │ 4.715 ns      │ 4.703 ns      │ 4.703 ns      │ 4.694 ns │ 100     │ 51200
   ├─ 8       55.58 ns      │ 113.8 ns      │ 59.21 ns      │ 58.74 ns      │ 58.11 ns │ 100     │ 6400
   ├─ 16      2.948 µs      │ 3.415 µs      │ 3.155 µs      │ 3.099 µs      │ 102.1 ns │ 100     │ 100
   ╰─ 32      4.015 ms      │ 6.609 ms      │ 4.448 ms      │ 4.627 ms      │ 535.4 µs │ 100     │ 100

[WhiteBlackGoose]

I guess it should work now maybe

[WhiteBlackGoose]

Maybe we should also cut off percentiles. Outliers skew the data too much, realistically we should perhaps keep the 1-99 percentiles, or even 5-95. But that's rather a future consideration?