Enable testing on drupal.org for patches.

Currently blocked on #2639408: Testbot installs the wrong dependency version

Currently the testsuite passes on php7(https://www.drupal.org/pift-ci-job/124864).

It doesn't pass on:
- php5.5 + sqlite (https://www.drupal.org/pift-ci-job/125166)
- php5.5 + mysql (https://www.drupal.org/pift-ci-job/125216)
- php5.6 + mysql (https://www.drupal.org/pift-ci-job/125222)

During the debugging of this issue; I also noticed that not all tests pass when ran trough the UI, even on php7. The test fails are different but they need more work.

I spent a couple of hours trying to debug the php5 fails but I didn't really get any closer to a solution. I set up the drupalci vagrant box locally and I can reproduce the php5 fails there.

We're expecting the result of the test (UrlIntegrationTest:97) to be "http://localhost/checkout/search-api-test-fulltext?y%5B0%5D=facet%7C%7Citem (y[0]=facet||item).

The 2 pipes are the seperator used in DummyQuery - a url processor defined in a test module. We've set the y as filter key earlier in the test.

    $edit = [
      'filterKey' => 'y',
      'urlProcessor' => 'dummy_query',
    ];
    $this->drupalPostForm(NULL, $edit, $this->t('Save'));

The result is "http://localhost/checkout/search-api-test-fulltext?f%5B0%5D=facet%3Aitem" (f[0]=facet:item). This is as if there is the save of the form in the test didn't work. But we have a specific test for saving the form that does work (it is green - also on php5).

I don't understand why this test doesn't work. It should.

So I wrote out a lot more specific in the test what we're actually trying to test. However there are still unexplainable things going on for php5.
At least the test is more explicit in what it's doing.

So, I was doubting that the actual saving of this might've been the issue. That's why I expanded the FacetSourceTest. See attached patch.

This verifies that the saving of the facet source is not what's going wrong.

Uploading the right patch now :)

Looks like the failures in travis are different to the failures here, because this issue didn't give me enough headaches already as it was.

Rebase.

We relied on some search api views that had the same path, see #2654860: Test views have the same path.. We now ship our own view (just one, not 2).

Many, many thanks to @Nick_vh for helping me out in debugging this issue.