Need a more robust way of handling exceptions from Firebase

[grodin]

Firebase throws exceptions for many things, often as way of signalling something non-exceptional. We need a more robust strategy for handling these exceptions, particularly as different Firebase components use exceptions in different ways.

For example, FirebaseAuthException has a number of different subclasses for different exceptional events, whereas FIrebaseFirestoreException is a single class with an enum field to tell you what kind of exception it is.

A first sketch for the auth "module" (not actually a separate module yet) might be something like

typealias FirebaseAuthServiceExceptionHandler = (FirebaseAuthException) -> FirebaseAuthServiceError

sealed interface FirebaseAuthServiceError

typealias FirebaseAuthServiceErrorHandler = (FirebaseAuthServiceError) -> WorkflowAction<AuthProps, AuthState, AuthOutput>

The first typealias would be used in FirebaseAuthService and the second in ActualAuthWorkflow.

It's likely that we also need a strategy for sending notifications back up the workflow tree. For example, if an exception is thrown because the user has become unauthenticated, we need that to go all the way back up to the root workflow so the user can log back in.

An enhancement to this could be to include the state of each workflow on the way up the tree so that once the user logs in again they can be returned to wherever they were rather than manually navigating back.

[grodin]

Related: #70

[grodin]

The first step to this is to stop using runSuspendCatching() and instead use an explicit try { } catch { } block. It's probably worth at least building something small to at least always rethrow CancellationException for coroutines.