Implement function to extract serialization for the events from a Projector

[forestphoenix]

See Issue #2

[forestphoenix]

Part of this was done in #4, but there is still need for a structure that can be expected to scale reasonably with many event types.

In addition, it would be nice if duplicate types (i.e. two deserializers for MyEvent) were reported somehow, ideally at compile time. --> Check OverloadedLists for this

[forestphoenix]

Hrm, when we re-use the projector to look the deserialization up, the risk of duplicate types will be very slim.

What is going to be much more interesting is how the deserialization of the projections can be found.

[forestphoenix]

The deserialization on the projections could be done when reading the projection. This would be an effective means to guarantee the deserialization is present when reading a projection. It would also remain an effective method when projections are stored/cached on a medium that cannot be deserialized entirely (such as a redis cluster)

However, this would pose a disadvantage in

• Finding corrupted data. The deserialization could only happen on demand, delaying the detection of coorrupt data (or defective update conversions)
• The elegance of the event store. The map holding the projections would have to be much more intricate if it also stored semi-serialized data

[forestphoenix]

Current plan of action:

• Proceed to extract deserialization for the events from the projector
• a cache would deserialize projections when they are read