Create a Git repository abstraction with just what we need.

This will be continued with a Book class holding/storing the info
about the whole book, its size and its authors.

Have dedicated objects:
- Repo
- Author
- Page
- Commit
handle all the data and responsibilities.

This commit creates the Page objects within the on_files
MkDocs event.

As a result the lines/contribution statistics for the whole book
are now available from within *every* Markdown page.

# Conflicts:
#	mkdocs_git_authors_plugin/plugin.py
#	mkdocs_git_authors_plugin/util.py

The 00000 "commit" shown by git blame for uncommitted lines
caused the plugin to crash when trying to execute "git show" on it.
Therefore uncommitted stuff is attributed to a fake author whose
display characteristics are configurable.

Inserting {{ git_authors_list }} in any Markdown file
will insert a list of all authors along with their contribution
to the whole site.

Adds configuration options:
- show_lines (default: false)
  also show the line count in the list
  (but not on a page's summary)
- label_lines (for localization)

TODO: Provide *real* configurability of the resulting HTML

Adds configuration options
- sort_by (choice: 'name', 'contribution')
- sort_reverse (bool, default: False)

There may be users not interested in counting empty lines
as content.

We don't have to go through the member accessor function
when we have the repo reference as a function argument.

This method is only used once and can easily be avoided by
directly passing the sha argument to Commit._populate.

Incidentally: Move the handling of uncommitted lines
*into* Commit._populate

This is not related to the actual object but a simple conversion
function and therefore an optimal candidate for a generic "util" module

The previous name raised concerns about users being worried
"commit" might be used as a verb (i.e. creating commits in the
actual repository).

The configuration variables intended for localization have to
be removed because the functionality will at one point be
handled by localization. See timvink#14

This is necessary to get a hold on the whole element
through CSS, for example to make it invisible (in certain
contexts).

It should be clear to any future contributor that when the
on_page event stage is reached all pages have already been
parsed and the repo-wide statistics are available.

This is to separate concerns: Move the formatting of the
list of authors to the repo not in the Git classes but separately.

Closes timvink#20

While parsing the --porcelain output of git blame is more complicated
than that of git blame -lts

* it should be considered more robust
* it provides substantially more information:
  - we can get all information about the commits that we need
  - this makes it obsolete to call `git show` on the commits

The format of the commit timestamp is different in this command,
therfore this had to modify the datetime processing functions,
which were moved to util.py at the same time.