This is follow-up to 5796a17a (core_test: Add test to make sure that
wendelin.core basically works; nexedi/erp5!1429).

In that commit it was said that testWendelinCore

    "currently passes with wendelin.core 1, which is the default.
     It also passes as live test with wendelin.core 2.
     However with wendelin.core 2 it currently fails when run on testnodes
     ...
     because we need to amend ERP5 test driver

     1. to run tests on a real storage instead of in-RAM Mapping Storage(*), and
     2. to spawn WCFS server for each such storage."

This patch addresses that latter problem to run testWendelinCore under
testnode infrastructure.

@rafael and @jerome suggested that we can force a test to be run on a
real storage via `runUnitTest --load --save` or via `--activity_node=n`.

@rafael also suggested not to generally change the testing driver, but instead
make step-by-step progress and first tag each test that uses wendelin.core with an
option. Let's go this way now: runUnitTest/custom_zodb are taught to launch
WCFS server if wendelin.core usage is requested and software is built with
wendelin.core 2.

With both changes combined testWendelinCore should now pass OK when run
on a testnode with both wendelin.core 1 and wendelin.core 2.

This patch is based on a draft patch by @rafael: https://lab.nexedi.com/rafael/erp5/commit/14e3a777.

This patch also relies on recent wendelin.core 2 wcfs.py rework which
exposed functionality to start WCFS server and to further control it:
kirr/wendelin.core@5bfa8cf8.

/cc @tomo, @romain, @jerome, @seb