Analytics | Create a new dedicated posthog/sentry event that better fit our needs

The current error event design is a bit flawed.
Grace period is too simple.
It reports UISIs errors for historical messages \(the keys was never distributed to you\)
Decryption has a lifecycle, it's possible to receive the keys after the message. We should reflect that in the metric, it has a start/finish \(time to decrypt\). The decryption can have several endings \(key withheld, from backup, via forward, or none\)

Start by locating the current error event and its PostHog/Sentry instrumentation. Define the event lifecycle for historical messages, delayed keys, and decryption outcomes, including start and finish timing; done means the new metrics reflect these cases consistently.

Written by the indexing model from the issue text.