Syslog memory usage fix

[alesnovak-s1]

After several attempts to make syslog monitor threading robust and memory efficient, I decided to implement a simple solution of bounded ThreadingMixin. A better candidate might seem to be a ThreadPoolExecutor, but it proved to be very memory hungry.

The reason for using threads in the first place is:

• UDP - avoid loosing too many sockets
• TCP - avoid blocking on unfinished and not yet closed requests

The ideal solution would be to get right of our rigid in-house implementation based on BaseServer and etither

• use socket.poll() on all open sockets or
• use async library for hansling sockets, i.e. tornado. But such a big refactoring is not a priority at the moment.

[github-actions[bot]]

Test Results

     20 files  ±0       20 suites  ±0   30m 40s :stopwatch: - 1m 12s 1 489 tests +1  1 469 :heavy_check_mark: +1    20 :zzz: ±0  0 :x: ±0  7 116 runs  +5  6 889 :heavy_check_mark: +3  227 :zzz: +2  0 :x: ±0 

Results for commit d9ac1355. ± Comparison against base commit 2abc1ca2.

:recycle: This comment has been updated with latest results.