feat(streamer): Replicate journal items from different fiber

[mkaruza]

No description provided.

[mkaruza]

Create additional fiber in JournalStreamer that will be used for sending journal items for replication. Fiber waits with baseline of 10000 us but if there is influx of journal items that period will be shorter by factor of 2.

Testing with dfly_bench

Both master and replicate nodes were running locally o 6 proactor threads and 8GB of RAM and dfly_bench was running as:

./dfly_bench --qps -100 --pipeline 150 -test_time 60 --ratio=1:0 -d 32 --proactor_threads=2 -c 5

~ NO REPLICATION (single node) ~

Total time: 5m0.010367772s. Overall number of requests: 45003000, QPS: 150010, P99 lat: 3497.56us
Latency summary, all times are in usec:
Count: 45003000 Average: 1844.1646  StdDev: 534958.76
Min: 10.0000  Median: 1791.4340  Max: 20280.0000

~ BASELINE REPLICATION ~

Total time: 5m0.010323251s. Overall number of requests: 45003000, QPS: 150010, P99 lat: 7763.2us
Latency summary, all times are in usec:
Count: 45003000 Average: 2846.6449  StdDev: 909460.22
Min: 17.0000  Median: 2605.8893  Max: 57202.0000

~ SPSC QUEUE REPLICATION ~

Total time: 5m0.01019038s. Overall number of requests: 45003000, QPS: 150010, P99 lat: 3551.86us
Latency summary, all times are in usec:
Count: 45003000 Average: 1622.9450  StdDev: 392480.80
Min: 11.0000  Median: 1565.2629  Max: 12383.0000

PROFILING

	MainLoop
NO REPLICATION	3.26%
BASELINE REPLICATION	9.10%
SPSC REPLICATION	6.38%

////

For more concrete test results there will be more testing needed.

[mkaruza]

Closing this draft. As discussed, replication improvement changes would be done gradually with few PRs.