feature(crypto): Add support for master key local pinning

[BillCarsonFr]

part of invisible crypto, follow up of #3607
Alternative fix for #3564
Alternative PR #3610

Add the capability to locally pin a public MSK for a ReadOnlyUserIdentity.
The first time an identity is seen for a user, the msk is pinned. Pin violation will be reported when an identity is rotated.

This PR only adds support for pinning, support for serialization/migration, persistance.
This could be used later by other PRs to report specific errors or show pinning violation to users.

Note about verification and pinning.
As part of this PR, if a new identity is detected it will still be seen as a pinning violation even if the new identity is signed by our usk. But the UserIdentity::has_identity_mismatch() will be ok. There is a pinning violation, but the new identity is verified and verification has priority.
That is to say that there is no auto-pinning if verified so far. To be discussed if we want it later

• Public API changes documented in changelogs (optional)

Signed-off-by:

[codecov]

Codecov Report

Attention: Patch coverage is 95.55556% with 2 lines in your changes missing coverage. Please review.

Project coverage is 83.98%. Comparing base (925c5b2) to head (42aae6a).
Report is 77 commits behind head on main.

Files	Patch %	Lines
crates/matrix-sdk-crypto/src/identities/user.rs	95.55%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3639      +/-   ##
==========================================
- Coverage   84.02%   83.98%   -0.05%     
==========================================
  Files         260      260              
  Lines       26712    26739      +27     
==========================================
+ Hits        22446    22456      +10     
- Misses       4266     4283      +17     

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

[richvdh]

This looks broadly sensible to me, though I'd defer to @poljar on this.

There are a bunch of documentation nits I'd like to see cleaned up, but let's agree on the basic structure first

[poljar]

I think the approach looks good, I left a bunch of nits though. I'll let @dkasak double check things.

[poljar]

Left a tiny nit, but I think that this is fine.

[dkasak]

I've left a bunch of wording suggestions and would advise changing the has_identity_mismatch method name to identity_needs_user_approval as discussed, but otherwise this seems good to me, so leaving an Approve for when suggestions are handled.