Split MockContext into Standalone and multi-group

[ahsimb]

Fixes #33

Created two classes MockContext and Standalone mock context instead of one MockContext. The interface of the MockContext is preserved. The current emit output of either of the classes can be accessed through the output property.

[tkilias]

Preliminary review. Had not yet time to check the tests.

[tkilias]

@Nicoretti suggested that we implement a generator which returns the column values and we collect these values in a list and compare this list at the end with an assert. Furthermore, we shouldn't test internal stuff like _current.
And, I thought about _next_group and _output_groups, I think, we can make them public, because the UDF will use the Interface UDFContext and not the MockContext.

def generator(ctx):
  while True:
    if not ctx._next_group():
      break
    while True:
          yield ctx.t2
          if not ctx.next():
            break

result=list(generator(context_set_emits))
assert result==['cat', 'dog', ...]

[ahsimb]

We don't want to put both groups in one stream. We could create a jagged array instead. But I am not in favour of having an extra piece of code in a test. I think testing should be straightforward if possible.

I should perhaps remove the checks of "private" variables, e.g. _current_context. On that note, I think we should make the _next_group and the _output_groups public. That would require some changes upstream, e.g. in UDFMockExecutor. Would you agree?

[tkilias]

agree with making _next_group and _output_groups public, because the class is anyway usually used through the Executor

and yes please remove _current_context and so on

[tkilias]

Regarding the remaining issue, let me think. about it.

[tkilias]

That with the jagged array sounds ok to me.

[tkilias]

I think, we should make this a independent function and not a static method