(REF) WebsiteTest - Mitigate flaky failures
Overview
--------
In recent days, api_v3_WebsiteTest has emitted sporadic failures like this:
```
api_v3_WebsiteTest::testDeleteWebsite with data set #0 (3)
Failed asserting that 3 matches expected 0.
/home/jenkins/bknix-max/build/build-2/web/sites/all/modules/civicrm/tests/phpunit/api/v3/WebsiteTest.php:75
/home/jenkins/bknix-max/build/build-2/web/sites/all/modules/civicrm/tests/phpunit/CiviTest/CiviUnitTestCase.php:209
/home/jenkins/bknix-max/extern/phpunit7/phpunit7.phar:615
```
and
```
api_v3_WebsiteTest::testDeleteWebsiteInvalid with data set #0 (3)
Failed asserting that 4 matches expected 1.
/home/jenkins/bknix-max/build/build-2/web/sites/all/modules/civicrm/tests/phpunit/api/v3/WebsiteTest.php:88
/home/jenkins/bknix-max/build/build-2/web/sites/all/modules/civicrm/tests/phpunit/CiviTest/CiviUnitTestCase.php:209
/home/jenkins/bknix-max/extern/phpunit7/phpunit7.phar:615
```
These failures do not reproduce for me in isolation.
Before
------
Both the failing assertions make an implicit assumption that the baseline content of `civicrm_website` is empty.
After
-----
The failing assertions use an explicit baseline (`$beforeCount`).
Comments
--------
The test failures are sporadic and only seem to seem occur when run in the full suite.
My theory is that something else is leaking `civicrm_website` records;
however, it's hard to track that down amidst a full suite (when the full
suite takes so long to execute). Therefore, I cannot be certain that this
is actually fixes the problem. However, this really just tightens up the
assumptions of the test - as long as it passes the PR tests, it should be
safe to merge and then watch in the `CiviCRM-Core-Matrix`.