DurableOrchestrationClient.getStatusAll fails if there is a terminated orchestration instance

*From an internal CRI -- 129336626*

> [PROBLEM DESCRIPTION]
>
> This issue is consistently reproducible. For a JavaScript durable function, if DurableOrchestrationClient.getStatusAll is called while there is a terminated orchestration instance, the call always fails. Sample error by log streaming (in instances.forEach, instances is what's returned by getStatusAll):
> ```
> [Error] Executed 'Functions.DurableFunctionsHttpStart' (Failed, Id=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx)Result: FailureException: TypeError: instances.forEach is not a functionStack: TypeError: instances.forEach is not a functionat module.exports (D:\home\site\wwwroot\DurableFunctionsHttpStart\index.js:45:23)at at process._tickCallback (internal/process/next_tick.js:188:7)
> ```
>
> [TROUBLESHOOTING STEPS]
>
> For a terminated orchestration instance, the Output column in the table DurableFunctionsHubInstances contains undefined. If the value is manually changed to a string value or to null, getStatusAll succeeds. Since undefined is not a valid JSON value, it seems some serialization operation is failing.
>
> This issue doesn't show up for C# durable functions.

The repro entry point is DurableOrchestrationClient.getStatusAll, with the failing consumer shown at index.js:45:23; start by tracing how a terminated instance's undefined Output is serialized from DurableFunctionsHubInstances. Done means getStatusAll returns a usable status collection for terminated instances, with regression coverage for that case.

Written by the indexing model from the issue text.