refactor: break down gbq.py file to several smaller ones

[sycai]

The intention of this PR is to re-organize the code to make it easier to implement the dry_run feature: #585

The file gbq.py contains too many lines which makes it hard for code navigations.

Change summary:

• GbqConnector and its helper functions are relocated from gbq.py to a new file gbq_connector.py
• Exceptions defined in gbq.py are relocated to exceptions.py file.
• Context class and context global instance are relocated from gbq.py to a new file context.py.

These items are all imported back to the gbq.py file to avoid breaking imports of our downstream products.

[tswast]

Let's be explicit about what aliases we are keeping around with __all__ + a comment explaining why these "unused" aliases are here.

[sycai]

__all__ may give us extra burden due to potential star imports. I used # noqa to mark all unused imports instead.

[tswast]

If we're refactoring, I wonder if we actually want GbqConnector at all for our main implementation? It seems like one of those "you don't need a class" situations https://www.youtube.com/watch?v=o9pEzgHorH0 since we aren't actually benefiting from any object oriented isms here.

[sycai]

As discussed offline. We won't address this issue for now.

That said, GbqConnector does seem like an over-engineered OOP class.