Allow GetWorkStatus to report NotStarted

[hendrikdevloed]

GetWorkStatus() throws an exception when the grain has not been started yet, versus returning NotStarted.

if (_status == SyncWorkStatus.NotStarted)
{
    Logger.LogError("{Method} was in a status of {WorkStatus}", nameof(GetWorkStatus), SyncWorkStatus.NotStarted);
    DeactivateOnIdle();
    throw new InvalidStateException(_status);
}
return Task.FromResult(_status);

In my use case I intended to start a long-running task (in my project a sequenceFactoryGrain eventually creates a sequenceGrain, working as a long-running task) in an idempotent way, like

ISequenceGrain sequenceGrain = await sequenceFactoryGrain.GetWorkStatus() switch
{
    SyncWorkStatus.NotStarted => await sequenceFactoryGrain.StartWorkAndPollUntilResult(sequence),
    SyncWorkStatus.Running => await sequenceFactoryGrain.ContinueWorkAndPollUntilResult(sequence),
    SyncWorkStatus.Completed => (await sequenceFactoryGrain.GetResult())!,
    SyncWorkStatus.Faulted => throw (await sequenceFactoryGrain.GetException())!,
    _ => throw new NotImplementedException()
};

but was surprised by the exception thrown when trying to check the current state.

My intent: If some external non-Orleans API call tries to start the factory to get a grain, but if somehow there is a timeout/cancellation/... and the API retries, I should be able to find out that the sequenceFactoryGrain is already processing and just tag along waiting for the long-running call to complete.
(ContinueWorkAndPollUntilResult is the same code as StartWorkAndPollUntilResult but without the Start call)

Originally posted by @hendrikdevloed in #55 (comment)

I'll make the modification to allow retrieving GetWorkStatus() and set up a pull request.