This is to lower the burden on implementers. The previous
approach (requiring a Page and Iterator subclass) ended
up causing lots of copy-pasta docstrings that were just
a distraction.

Follow up to googleapis#2531.

The previous implementation may catch users off guard since
the iterator.page access may also update the value before
access.

In addition, this PR removed the _update_page() / next_page()
subclass behavior in _BlobIterator. Over-riding that method
was never intended. Instead makes a non-public class attribute
_PAGE_CLASS that can be replaced with Page subclasses. This
can be revisited if more implementations require custom
behavior on Page creation / Page.__init__.

Also renaming next_page() to update_page() on Iterator and dropping
any return value from that method. Also throwing an AttributeError
if the page is unset on @Property access.

This is because Iterator.page combined with Iterator.update_page()
can provide the same thing and has_next_page() is really an
implementation detail.

Done via:

$ git grep -l has_next_page |
> xargs sed -i s/has_next_page/_has_next_page/g