Server validation incorrectly passes with various unimplemented features

[ml-evs]

Just noticed that validation passes with some "lucky" choices of missing features, e.g.:

• If a server does not support page_limit (and silently ignores it), then the validator for response_fields will pass (as there is a check is for len(results) == page_limit, not len(results) > 0)
• If a server does not support page_limit and coincidentally uses the same default page_limit as the the validator, then it will pass the page limit tests
• If a server does not support page_offset then it can get away with not supporting filter (!) as the same "random" structure is chosen for the archetypal filter validation.
• If a server does not support filter at all, thus returning all results to all queries, all filter tests would pass as the validator treated more_data_available as a sign that the filter was too large to be useful for testing