dnsdist: Create TCP worker threads before acceptors ones

[rgacogne]

Short description

Otherwise we might crash when the first TCP query arrives because we are dividing by zero when trying to select a worker.
This PR also adds an explicit check to turn the crash into an exception if we ever manage to break that again.

Checklist

I have:

• read the CONTRIBUTING.md document
• compiled this code
• tested this code
• included documentation (including possible behaviour changes)
• documented the code
• added or modified regression test(s)
• added or modified unit test(s)

[rgacogne]

Unrelated failure in remotebackend-pipe tests, with not much to investigate:

basic-a-resolution
--- ../modules/remotebackend/regression-tests/basic-a-resolution/expected_result	2021-02-16 10:46:16.000000000 +0000
+++ ../modules/remotebackend/regression-tests/basic-a-resolution/real_result	2021-02-16 11:10:45.582292569 +0000
@@ -1,3 +1,2 @@
-0	outpost.example.com.	IN	A	120	192.168.2.1
 Rcode: 0 (No Error), RD: 0, QR: 1, TC: 0, AA: 1, opcode: 0
 Reply to question for qname='outpost.example.com.', qtype=A
./start-test-stop: line 192: kill: (12520) - No such process

Exited with code exit status 1

CircleCI received exit code 1

[Habbie]

Unrelated failure in remotebackend-pipe tests, with not much to investigate:

I've seen similar, very seldom, failures in the remotebackend. I haven't found why yet.