* Prefer ids over "foreign" domain objects.
* Consistently use final keyword for implementation parameters and
  variables.
* As a consequence, session id is used instead of the object as cache
  key for `skillquestions`, `lecturequestions`, `preparationquestions`
  and`flashcardquestions` caches.
* Rename occurances of `questionId` to `contentId`.

Added new service for comments.

* Moved classes from `de.thm.arsnova` root package.
* New util package
* Do not start interface names with `I`. Appended `Impl` to the
  implementing class instead.
* Renamed learning progress classes to score and move them from
  `.domain` to a new `.services.score` package.
* Do not capitalize acronyms in class names:
    * CASLogoutSuccessHandler -> CasLogoutSuccessHandler
    * SocketIO* -> Socketio*
* Renamed `.socket` package to `.websocket`

Our version of CouchDB4J has been updated to fix and improve type
handling for database keys. Strings are no longer converted implicitly
to integers or booleans based on their content. This lead to bugs if key
and property values were compared.

Fix registration of Bean for socket event handling

See merge request !70

Fixes a regression introduced by commit
2ecbfec8.

Fix design doc creation

See merge request !69

Authentication is needed if an CouchDB admin user exists for design doc
creation at startup.

Closes GH-6.

Migration to Ektorp and refactoring of the database layer

A summary of the changes:
* CouchDB4J has been replaced by Ektorp.
* CouchDBDao has been split up into multiple repositories: One for each
  entity.
* Jackson is used instead of JSON-lib for POJO <-> JSON
  (de)serialization.
    * Jackson's `@JsonView`'s are used to mark which properties are
      (de)serialized for the API and/or CouchDB
    * A lot of persistence code could be simplified since serialization
      is handled automatically by the `ObjectMapper`.
* Entities and related Classes have been renamed:
    * `Content` <- `Question` / `SkillQuestion` / `LecturerQuestion`
    * `Comment` <- `FeedbackQuestion` / `AudienceQuestion`
* CouchDB design docs can now be automatically created by the backend if
  necessary. They are included as JavaScript code which is parsed and
  transformed to JSON at startup using Java's `ScriptEngine`.
* `MediaType`s for API versioning have been introduced:
    * `vnd.de.thm.arsnova.v2+json`: the format as used in ARSnova 2.x
    * `vnd.de.thm.arsnova.v3+json`:
        * ISO 8601 dates instead of timestamps
        * `id` instead of `_id`, `revision` instead of `_rev`
* `api.indent-response-body` can be set in configuration for pretty
  responses for API debugging.
* Empty (e.g. `null`) entity properties are ignored for serialization

See merge request !68

Mosts of the tests were coupled with the old database layer.
Additionally, a lot of them were testing simple CRUD operations intead
of domain logic. Meaningful tests with proper database mocking have to
be written at a later time.

Previously, the dependency was indirectly included via CouchDB4j.

Code for the following domains has been migrated:
* VisitedSession
* MotdList
* Statistics
* LearingProgress

(Lecturer/Skill)Question has been renamed to Content. Method names have
not been touched yet to ease reviewing the changes.

InterposedQuestion has been renamed to Comment. Method names have not
been touched yet to ease reviewing the changes.

LoggedIn documents are also handled by the SessionRepository for now.

Java's ScriptEnging is used to transform design doc JS objects into
valid JSON. A custom CouchDbConnector has been implemented to handle
this transformation and create the design docs in the database at
startup.

Priortizing v3 caused problems for v2 clients on POST requests.