Extend skeleton for LOM secure aggregation training

[mvesin]

As a developer I want to extend researcher, node and messages classes "skeleton" so that I can trigger a training with secure aggregation (choosing scheme from Joye Libert and LOM) with the Experiment API from a notebook, with a dummy payload (not yet really doing encryption, or saving/reading secagg elements to/from database).

The output of this task a a "skeleton" implementation where we can plug the bits: LOM encrypt/decrypt, DH negotiation payload, secagg elements save/read to/from database, etc.

• extend researcher SecureAggregation class hierarchy

• extend researcher SecaggContext class hierarchy

• adapt researcher breakpoint

• adapt/extend message format for TrainRequest (add secagg scheme, specific ID(s) per scheme)

• adapt node side Round

• create SecaggLomCrypter with dummy payload (where LOM code will be inserted later)