feat(ffi): add support for starting and responding to user verification requests

[stefanceriu]

This patch expands on the existing device verification flows to add support for verifying users as well.

It does so by:

• moving incoming request detection to the Client level so that it works for both to-device events as well as timeline ones. The client then forwards the data through the generic process_incoming_verification_request method
• exposing a new request_user_verification method next to the (renamed) request_device_verification one on the SessionVerificationController
• consolidating subscription logic within set_ongoing_verification_request
• overall the mecanisms are very similar which should make it easy to adopt in the final clients

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.73%. Comparing base (d5d9898) to head (6ce266a).
Report is 46 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4618   +/-   ##
=======================================
  Coverage   85.73%   85.73%           
=======================================
  Files         291      291           
  Lines       33323    33325    +2     
=======================================
+ Hits        28570    28572    +2     
  Misses       4753     4753           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[poljar]

Looks good. I left a couple of nits and it's quite sad that we can't, instead of adding event handlers here, have a higher level stream in the SDK which gives you the Verification object once one gets created.

Maybe one of these days when the OlmMachine doesn't get permanently killed and resurrected. 🥲

[poljar]

Is there a point in removing these? Perhaps it would make sense to have a helper that always applies this context to the UserId parsing step instead of just removing this?

[stefanceriu]

I just looked at where UserId::parse was used and noticed there's significantly more without a context than with. Figured I might as well make it consistent.

I'm not sure it's worth wrapping it, I don't think I've ever seen one of these in practice and even if it were to happen the underlying ruma error is clear enough.