… branches with different crates)

…instead of libparsec_testbed) to simplify dependencies on user crate

…torage implementation

…ltiple testbed servers

… tokio

…on_speculative_init

… .json/.json5 files

…c_protocol

…ration of python bindings), improve tests

… factorization

…p by correct dump on each cmd's Req, rework protocol tests

…atibility.py)

`messaget_get` command had an invalid type for it nested type
`Message.count`, this is purely cosmetic though given `Size` and
`Index` (old and new types used) are actually handled in the
same fashion by the serialization format.

…est_load_template` (given it could be used for pythen tests)

- Remove `wait` param from `events_listen` (uneeded given only SSE
  should have access to this command in APIv4)
- Add `not_available` status to `events_listen` for when it is used
  outside of SSE
- Modify `MessageReceived` event type in `events_listen` to return the
  message body (given it is small and it saves a roundtrip)
- Add `CertificatesUpdated` event type to `events_listen`
- remove `events_subscribe`,`human_find`,`realm_get_role_certificates`/`user_get`
  (replaced by the `certificate_get` command & `CertificatesUpdated` event)
- add `bad_timestamp`&`require_greater_timestamp` response statuses for all
    commands doing modications

…tificate` certif

…ings

… APIv3 (removal of events_subscribe & new rep statuses)

- SSE requires a correct request payload (i.e. `events_listen` command) this
  make possible future changes where the client could optin/out of events types
- SSE uses POST verb (instead of GET), this is required given previous point
  and provide consistency with the default RPC
- SSE connection is correctly closed if the user is no longer legitimate
 (e.g. in case of user revocation)
- APIv3 is not allow on SSE, this simplify the implementation and has no use anyway

…g device_id/user_id

…her class

…ery misleading)

…mand

… stress this has security implication if not used properly !

…essage

…nd message

…avoid needing long and combersome error names

…nnection crate !

…xplicit expected errors

…y checks

…c_client)

…tion with UserUpdateCertificate

…ed in the Rust client)

- add UserUpdateCertificate
- improve IndexInt support
- support common unsecure_load for SequesterServiceCertificate
- rename profile -> initial_profile in UserCertificate

- Add bad_timestamp error status in device_create
- replace timestamp by index in events_listen & certificate_get
- Return certificate index in vlob_read
- rename fields in message_get response for consistency with events_listen

…:start

…nit variants

Core is currently broken and planned to be replaced by the Rust client,
so better remove it now to cleanup the codebase ^^

…small data and add related tests

As enum, DeviceFileType is only serialized as a string (and given packing doesn't
support this custom extension, it should have never been used anyway
)

ApiVersion field is only used in the hanshake protocol, and is serialized
as a tuple.
Funny enough, the only place where this custom extension was used was in
the tests (so the generated payload was incorrect, but it worked anyway
because once deserialized ApiVersion is a tuple so it could be deserialize
again by the actual desialization code !)

…rypto