Integrate PSI with workers

[TTitcombe]

Question

The current PyVertical process requires a user to collect ids from the data holders to compute the set intersection. The data holders must then sort the remaining IDs they hold in the same format. We should integrate the PSI process more closely into vertically partitioned datasets. How we do this is an open question. We must consider:

• data holders to agree on a process for sorting IDs
• method for data holders to communicate IDs to a third-party for set intersection to be computed
• method to confirm that data holder IDs are/can become strings
  • PSi requires strings

Additional context

Part of #28

[TTitcombe]

This functionality will be developed in an experimental notebook in the PySyft repo, as we need to make use of syft 0.3.0 features

[TTitcombe]

workers not included in syft 0.3