Currently, all error boundaries close audio and Apollo connections once
an error is caught. This is not the correct behavior as not all error
boundaries are critical, e.g.: the presentation crashing should _not_
break the whole client. It also deviates from how error boundaries
worked in 2.7
Add a new prop to the ErrorBoundary/LocatedErrorBoundary components
called isCritical that flags an error boundary instance as critical. If
true, it'll close Apollo/audio. The default behavior is
isCritical=false, and the only critical error boundaries are the ones
located in the app's root (/client/main.tsx).
The ErrorBoundary component has a generic log message that is not
overridable and will aggregate errors from different components. The
ideal scenario is that log tags are configurable so that errors can be
filtered for in log post-processing.
Add the logMetadata: { logCode, logMessage } prop to the ErrorBoundary
component so that log tags are configurable. Default tags are
maintained.
Add specific log metadatas for the client startup error boundary
(logCode: 'app_startup_crash') and the user connection error boundary
(logCode: 'app_crash').
