Fix session creation

[ebrahimebrahim]

There are problems with db.add_session:

• Session.to_dict fails when there is no volume (recall from 87c4482d37a7e722656aa3c21a04784a448204d4 that we wanted to phase out volume in favor of volume_id – so apparently that transition needs to be taken to completion).
• Serializing a session results in duplication of the entire transducer geometry data. So if you start with a session that defines a transducer by referencing its ID, then when you serialize that you end up with a session that defines a transducer by defining its geometry. The serialization/deserialization of Sessions needs to be cleaned up like we did for Protocols and Transducers in #85
  • UPDATE: There is now a new data format that changes the approach we need to take here. See #100:

    In the updated example database, the session.json files now include a reference to the trasducer_id used instead of directly storing all of the transducer information. from_dict in db.session needs to be updated to parse the transducer_id key instead of looking for the transducer key.

• There is no check that the subject ID inside the session json is the same as the the subject specified in the call to db.add_session

[ebrahimebrahim]

Do sessions contains a solution id or something like that? We might want to consider saving solutions as part of sessions, in which case this would partly depend on #103