Clarify failure modes around GetXxxContainerForTenantAsync

In https://github.com/marain-dotnet/Marain.UserNotifications/pull/112 it became apparent that we have a global lack of clarity around expected error handling behaviour in cases where the requested container cannot be created for some reason.

There are at least two different reasons this could happen:

1. transient infrastructure failure
2. incorrect configuration

The documentation for all of our tenanted storage providers says nothing about how to distinguish between these. This had percolated up the stack causing the same essential problem to become visible in the PR in `Marain.UserNotifications`. But that PR can't solve that problem on its own. It first needs to be addressed at the `Corvus.Tenancy` level, and once we've established the pattern here, that will have knock-on consequences for anything that consumes this.

Start by reviewing the behavior discussed in Marain.UserNotifications PR #112 and the GetXxxContainerForTenantAsync APIs in Corvus.Tenancy. Establish how transient infrastructure failures differ from incorrect configuration, then document the resulting pattern for tenanted storage providers and identify its consequences for consumers such as Marain.UserNotifications.

Written by the indexing model from the issue text.