chore: fix criterion benchmarks

[FrankReh]

Follow up to #151.

[FrankReh]

@Noah-Kennedy This is the kind of output I'm getting from the criterion benchmark with these changes. These times jive with what I was seeing with a different form of no_op benchmark I had experimented with before: that when there is little concurrency, the time per no_op is high, say 3.2 microseconds, but when there are lots of concurrent uring accesses, the amortized time per no_op is very low, about 332 nanoseconds on my slow Linux machine.

     Running benches/criterion/no_op.rs (target/release/deps/criterion_no_op-22257b96c5b4fa14)
no_op/concurrent tasks/1
                        time:   [3.2306 µs 3.2484 µs 3.2698 µs]
no_op/concurrent tasks/2
                        time:   [1.7880 µs 1.7974 µs 1.8086 µs]
no_op/concurrent tasks/5
                        time:   [877.98 ns 882.60 ns 888.23 ns]
no_op/concurrent tasks/10
                        time:   [574.71 ns 577.88 ns 581.69 ns]
no_op/concurrent tasks/100
                        time:   [324.33 ns 326.25 ns 328.63 ns]
no_op/concurrent tasks/1000
                        time:   [309.72 ns 311.48 ns 313.65 ns]
no_op/concurrent tasks/10000
                        time:   [465.57 ns 469.27 ns 473.69 ns]

[FrankReh]

I guess you have some issues with how I did it. I will have to pull your branch and see if you get the same kind of results, only better.