[quasimodo] Integrate mongodb world_state observations and enable querying individual databases

[nilsbore]

This PR integrates quasimodo with the world_state representation of the object search framework. It enables us to use the mongodb observations intermixed with the meta room ones. Further, it uses this representation to enable @jekekrantz to query for top matches among his built models to speed up matching of models in his database. This will be used to associate observations from @jayyoung 's object search system.

If @jekekrantz could test this before merging, that would be good. Note that when searching for objects in your models, you need to pass the flag query_msg.query_kind = quasimodo_msgs::retrieval_query::MONGODB_QUERY.
To add a model to the database, you need to have mongodb running, as well as rosrun quasimodo_object_search insert_object_server.py. You then need to pass a service according to this definition, in quasimodo_msgs/insert_model.srv:

uint64 INSERT=1
uint64 REMOVE=2
uint64 action
quasimodo_msgs/model model
string object_id

---
uint64 vocabulary_id
string object_id

This is currently announced at /insert_model_service. For INSERT, you only need to pass the model argument. You then get an object_id back that you need to keep track of. When you want to remove this model again, pass REMOVE together with the object_id.

@jayyoung If you have soma_manager/soma2_local.launch running together with quasimodo_object_search/create_object_search_observation.py, this code will automatically add new entries added to the world_state database. For querying these observations, the procedure is the same as above. ATM, you need a semanticMap database to train the underlying representation to be able to use this.

[nilsbore]

So this finally(!) passes. @RaresAmbrus OK to merge?

[nilsbore]

Ok, @RaresAmbrus , I think this is fine to merge now, I included some licensing text in the stopwatch project.