Create a simple_cloud fixture

[torcolvin]

In order to implement rosmar support into the e2e framework, it makes logical sense that couchbase_servers is nil. However, all tests that use configure_dataset require a CouchbaseCloud instance.

This PR only creates a simple_cloud convenience function to allow for a future rosmar upgrade, for all tests that can be run without an actual Couchbase Server instance (most of them).

[borrrden]

Is the goal here to be able to run all existing tests with rosmar? Because otherwise I think if you write a new test there is no requirement to use the cloud class. That class was there specifically because setting things up between server and sync gateway is finicky and messy. For rosmar you can just skip straight to SGW db PUT right?

[torcolvin]

Is the goal here to be able to run all existing tests with rosmar? Because otherwise I think if you write a new test there is no requirement to use the cloud class. That class was there specifically because setting things up between server and sync gateway is finicky and messy. For rosmar you can just skip straight to SGW db PUT right?

The goal is to run all the existing tests, the tests all run OK except those that really require a server (XDCR). If you'd rather that I do something hackier like putting:

  "couchbase-servers": [
    {
      "hostname": "fakelocalhost",
      "rosmar": true
    }
  ],

I can do that to make sure that it doesn't accidentally try to contact a couchbase server that might exist and thus not work correctly.

To me, these fixtures reduce LOC and complexity but I can understand why this isn't always that way.

Another way of thing about this instead of a fixture like this would be to replace the cloud fixture with a method on CBLPyTest.cloud(self) that does the same work as the test fixture.

I can infer rosmarness by asking Sync Gateway, so technically this doesn't even need to be present in the configuration file at all.

[borrrden]

If you are ready for this to be reviewed let me know again because the checks are failing

[torcolvin]

I put this up as an idea of what to do. The code that can go into production is in #375

I mostly want to know if you think this makes things better (or worse).