Gdocrevisions

Bugfixes for Document.at_revision() and .at_time() methods:
* convert filter to list (python3)
* apply each revision separately due to depreciated support for list in Revision.apply()

* Uses google-auth instead of oauth2client for google credentials (https://github.com/harvard-vpal/gdocrevisions/pull/6). See https://google-auth.readthedocs.io/en/latest/oauth2client-deprecation.html) for oauth2client depreciation notes.
* Removes suboperation class (speed up document parsing and reduce memory usage)
* Supports google docs using "suggestions" (track changes) (https://github.com/harvard-vpal/gdocrevisions/pull/7)
* Python 3 support (https://github.com/harvard-vpal/gdocrevisions/pull/8)
* Utilities for measuring GoogleDoc initialization running times (https://github.com/harvard-vpal/gdocrevisions/pull/9)
* Adds tests (https://github.com/harvard-vpal/gdocrevisions/pull/10)

* Memory optimization with python `__slots__`
* Undo functionality
* Optimized replay class
* Measure elapsed times for Document methods

* Introduces Suboperations for tracking single-character (or element) changes
* Single Content.apply() method replaces Content.apply_operation(), Content.apply_revisions(), etc

* Bugfix for docs with footnotes, by including EndOfBody dummy element type for separating document body from footnote text
* Create Character instances on insert operation initialization, rather than on apply
* Easier document initialization - support specifying service account keyfile location as opposed to creating and passing a credentials object to
* iter_operation methods on document/revision/operation objects for flattening operation tree and easier iteration through operations