The polling of drop flag and memory allocation method could be improved about udp-stream

[asuper0]

benchmark code here shows the improvement.

and the result:

     Running benches\benchmark_lib.rs (target\release\deps\benchmark_lib-36879d0bb7f740c7.exe)
Benchmarking sample-size-example/normal: Warming up for 3.0000 s
Warning: Unable to complete 10 samples in 5.0s. You may wish to increase target time to 121.1s.
sample-size-example/normal
                        time:   [11.078 s 11.456 s 11.868 s]
                        change: [-81.264% -80.491% -79.680%] (p = 0.00 < 0.05)
                        Performance has improved.

     Running benches\benchmark_optimize_buf.rs (target\release\deps\benchmark_optimize_buf-bfd7c03693bf79e1.exe)
Benchmarking sample-size-example/optimize buf: Warming up for 3.0000 s
Warning: Unable to complete 10 samples in 5.0s. You may wish to increase target time to 79.7s.
sample-size-example/optimize buf
                        time:   [8.5379 s 8.5776 s 8.6159 s]
                        change: [-80.839% -80.737% -80.642%] (p = 0.00 < 0.05)
                        Performance has improved.

     Running benches\benchmark_optimize_drop.rs (target\release\deps\benchmark_optimize_drop-9e46585be99cc901.exe)
Benchmarking sample-size-example/optimize drop: Warming up for 3.0000 s
Warning: Unable to complete 10 samples in 5.0s. You may wish to increase target time to 95.0s.
sample-size-example/optimize drop
                        time:   [9.3510 s 9.3941 s 9.4321 s]
                        change: [-82.663% -81.850% -81.082%] (p = 0.00 < 0.05)
                        Performance has improved.

The optimization of drop method reduce the time from 11.456s to 9.3941s, then the one of BytesMut reduce to 8.5776s.

[SajjadPourali]

Looks good, thank you very much.