Feature similarity groups

[gusmith]

This is a first step to have the similarity score output for multi-party: this PR modifies the output from

[
  [row_index_0, row_index_1, score],
  ...
]

to

[
  {'group': [[party_id_0, row_index_0], [party_id_1, row_index_1]], 'sim': score},
  ...
]

which follows the groups output format of the candidate pairs (i.e. a record is represented by two indices [party_id, row_index].

[wilko77]

I am concerned about the size of the output. By wrapping every sim score in a json object, we would almost double the amount of characters in the output.
I think we should go with something more streamlined. I made a suggestion in the comments, happy to discuss.

👍 for keeping the changelog up to date.

[wilko77]

just some minor comments.
But have a look at the permutation notebook test. Something is not right.

[gusmith]

The tutorials are currently faliing because of the newest release of clkhash which has some breaking changes.
The PR #465 is fixing them, but in progress.

[gusmith]

@wilko77 Since your review, I simply merged dev into this branch (resolving some conflicts in a tutorial), and added a note about the tutorials usage, mainly because the ones from this branch are NOT working with the currently deployed service (because of some breaking change in the code which has not been deployed).