Initial TCP high-water value is wrong

This is a minor glitch since TCP high-water, as the name suggests, cares about the highest value since restart... However, when the feature was still developed, the typical rndc status output for a dual-stack server listening on two addresses would be like this fresh after startup:

tcp clients: 2/150
TCP high-water: 2

This is correct.

However, after netmgr (!2528 (merged)) was merged, the output after startup is:

tcp clients: 2/150
TCP high-water: 0

Issuing a TCP query restores the proper value:

tcp clients: 2/150
TCP high-water: 3

But the initial count should still be fixed to prevent confusion. The tcp system test does not catch this because the system test framework uses TCP queries to verify that a given server is up and thus the counter is already correct by the time tests.sh is executed.