refactor: speed up IO

[crepererum]

1. refactor: use sync lock in request path

$ env TEST_INTEGRATION=1 KAFKA_CONNECT=localhost:9010 cargo bench --all-features --bench throughput
...

read/rskafka            time:   [7.1591 us 7.2529 us 7.3444 us]
                        change: [-33.377% -22.989% -10.606%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  1 (6.67%) high mild
  1 (6.67%) high severe

...

write_sequential/rskafka
                        time:   [335.19 us 347.14 us 366.57 us]
                        change: [-6.1186% +1.7049% +11.034%] (p = 0.71 > 0.05)
                        No change in performance detected.

...

write_parallel/rskafka  time:   [15.904 us 16.063 us 16.261 us]
                        change: [-4.8534% -2.3144% +1.2491%] (p = 0.16 > 0.05)
                        No change in performance detected.
Found 1 outliers among 15 measurements (6.67%)
  1 (6.67%) high severe

2. refactor: avoid thread/task bouncing when sending data

The problem with spawning tasks (or a MPSC construction) is that tokio
will likely execute the task in another thread and requires additional
futex syscalls for synchronization. When sending data we were wrapping
the sending task to prevent cancellation. However cancellation is a
rather unlikely event (unhappy path) and we should only pay for it when
it is really needed.

$ env TEST_INTEGRATION=1 KAFKA_CONNECT=localhost:9010 cargo bench --all-features --bench throughput

...

read/rskafka            time:   [8.9115 us 9.2918 us 9.6791 us]
                        change: [-4.0247% +3.9206% +12.188%] (p = 0.36 > 0.05)
                        No change in performance detected.

write_sequential/rskafka
                        time:   [279.21 us 292.64 us 304.67 us]
                        change: [-21.742% -16.312% -10.823%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) low severe
  2 (13.33%) low mild

...
write_parallel/rskafka  time:   [15.617 us 15.821 us 16.061 us]
                        change: [-3.4937% +0.0064% +4.3441%] (p = 0.99 > 0.05)
                        No change in performance detected.
Found 1 outliers among 15 measurements (6.67%)
  1 (6.67%) high severe