Fixes: #2200

Because it uses hard-coded ports, it often appears as failing in vtest
so now it should succeed if the address wasn't available.

Spotted by Martin, when a VMOD relying on varnish.m4 is built with
--prefix in the configure command-line, a pseudo hard-coded vmoddir
is assumed:

    $libdir/varnish/vmods

Instead, it now asks pkg-config for the vmoddir relative to the
libdir. It worked in all cases with upstream Varnish Cache, but
downstream modifications of the vmoddir would then be ignored.

Fixes #2197

Change some tests to use the new parameter and more coverage.

Comment it out for now and add a note for future reference.

the exit code test, than not doing so afterwards.

Not part of a release, so not using a deprecation message.

While at it, kill the `workdir` field that is neither used nor exposed,
and wrongfully based on the misleading ${pwd}. The overall documentation
also got a refresh, not just the new macros.

My aim is to have one u* vtc per tool.

Add a new counter, n_test_gunzip to track the former.

Also clarify %{X}i and %{X}o wrt multiple X headers.

Fixes #2190.

Have varnishtest test varnishtest (don't look!)

Fail out of http reception if the socket fails

it at the end of the run where it is far more likely to be noticed.