Minimized allocations

[frol]

The original results on my laptop:

runtime         time per
async-std       0.980
tokio           3.499
async-std       0.749
tokio           3.506
async-std       0.760
tokio           3.729
async-std       0.776
tokio           3.551
async-std       0.734
tokio           3.441
async-std       0.757
tokio           3.455

The new results:

runtime         time per
async-std       0.831
tokio           3.340
async-std       0.575
tokio           3.306
async-std       0.564
tokio           3.351
async-std       0.687
tokio           3.467
async-std       0.570
tokio           3.356
async-std       0.557
tokio           3.354

NOTE: It does not solve the oddity of tokio I observe in #3 (comment)