feat(session): Add unhandled session status type

[ryan953]

This is in service of this project: https://www.notion.so/sentry/RFC-Session-Status-unhandled-2308b10e4b5d801d9805fece61533ac5#2308b10e4b5d80e9b312ecbea27a4af3

TL/DR:
We're adding a new enum value to split up the existing crashed release health metric into two more specific metrics:

• Unhandled
• Crashed
  For cases like Flutter: a Dart error will not crash the running process, while a native error will. We want to be clearer about this when we report errors to sentry users.

This PR
This PR will allow us to properly count session and sessions envelopes that contain the new value. When a session envelope arrives that has the new value: status: unhandled we want the SessionLike metric to handle that the same way we handle status: abnormal or status: crashed. We'll subtract 1 from the errors count and return 1 from unhandled_count().

Similarly, for sessions envelopes we need to implement the unhandled_count() function.

Finally, we have to implement the if-statements that will get the new count from the session, session.unhandled_count() and push the value into the list.

Fixes REPLAY-549

[linear]

REPLAY-549 Update Relay to correctly count envelopes that contain the enum val

[Dav1dde]

We have some session integration tests in test_session.py, would be great if you can add/modify them to also test the new codepaths.

[ryan953]

@Dav1dde I don't think that test_session.py is testing any of the properties of the session envelope, it looks like it's only testing for the filter rules that accept/deny envelopes.

instead i'll add some tests at the bottom of session.rs and mod.rs. The existing tests are not comprehensive to test all the session state types, but i'll expand them as best i can.

[Dav1dde]

It would be good to still have a single integration test in Python, you're right it seems like the tests in test_session mostly focus on dropping sessions, we do have some that test metric conversions (when searching for usages for send_session) which can work as a template.

I generally want at least one integration test for every protocol change, just to show it works, especially since integration tests very easily can setup a 'Relay chain' to make sure, the protocol change also works when interacting with multiple different types of Relays. A Rust unittest, can unfortunately not do that.

On a side note, I am now seeing that these tests are pretty worthless, as Relay stopped supporting producing to a sessions topic over a year ago :(