[Benchmark] [Participant Simulation] Protocols Derandomization: Networking Layer Change

[SimonRastikian]

As described in #138 , derandomization aims at taking a snapshot of a (any) protocol run.
For this we require changing of the networking layer or the protocol running function to enable writing in an external (ephemeral) file all the data sent by the participants.

The snapshot will consist of a set of files, one per party, recording all the messages sent by the corresponding party and in order.
For example: taking the snapshot for participant Pj means dynamically creating a file sent_pj which will contain records in the following spirit:
Pj sent to P0:
Pj sent to P8:
Pj sent to P0: -------- yes this repetition could be that round 2 started before round 1 ends
Pj sent to P1:
etc...

To preserve the order, every sent function should take hold of a lock to prevent any concurrent writing in the file.

Acceptance Criteria:

• Create one file per participant
• Append sent data per participant in the file in a precise encoding manner
• Create a lock system for the files
• Implement decoding function for the files that is able to read the files and properly collect the data.
• Test encode/decode

[SimonRastikian]

The implementation would encode differently:
Suppose P0 sent to everyone a message m.
The encoding would generate a file P0.raw containing the identities of each participant in the order of the sending e.g.:
P5 m.len() m P3 m.len() m P2 m.len() m P1 m.len() m ... etc,
All of these are written in raw bytes and not in a human readible manner

[SimonRastikian]

After a discussion with @netrome and @gilcu3, the decision was to perform the snapshots in-memory instead of files.

[SimonRastikian]

Acceptance Criteria:

• The snapshot has clean structures that define well the behavior.
• The snapshot in-memory still should be tested on a real Protocol run.