"tcp" system test fails intermittently and needs a cleanup

The TCP high-water tests added in !2425 (merged) tend to fail under high load on some operating systems (notably FreeBSD):

S:tcp:Wed Nov  6 22:28:05 UTC 2019
T:tcp:1:A
A:tcp:System test tcp
I:tcp:PORTRANGE:12600 - 12699
I:tcp:initialize counters
I:tcp:check TCP transport
I:tcp:check TCP forwarder
I:tcp:TCP high-water: check initial statistics (1)
I:tcp:TCP high-water: check value after some TCP connections are established (2)
I:tcp:TCP high-water: check value after some TCP connections are closed (3)
I:tcp:TCP high-water: ensure tcp-clients is an upper bound (4)
I:tcp:incorrect current TCP clients count: expected 8, got 17
I:tcp:incorrect TCP high-water value: expected 10, got 17
I:tcp:failed
I:tcp:exit status: 1
R:tcp:FAIL
E:tcp:Wed Nov  6 22:28:14 UTC 2019

Furthermore, the older checks in the tcp system test are not numbered, preventing forensic data from being available in case of intermittent failures.

While we are at it, we should also make sure that the aforementioned test employs set -e and that ShellCheck does not raise any warnings for it.