Forked Workflows call didComplete() early

When forking a Workflow the didComplete() is called as soon as the first branch completes. The fix that added didCompleteIfNotYet() guarantees that didComplete() is called only once, but that one time is when the first branch completes. Intuitively, I would expect the didComplete() to be called once all of the branches have completed, but that could be a bad assumption.

Related to #211 and the fix in #214

Example Test Case:

func test_workflowForkCompletionOrder() {
        let workflow = TestWorkflow()
        
        let emptyObservable = Observable.just(((), ()))
        _ = workflow
            .onStep { _ -> Observable<((), ())> in
                return emptyObservable
            }
            .onStep { _, _ -> Observable<((), ())> in
                let forkedStep: Step<(), (), ()>? = emptyObservable.fork(workflow)
                forkedStep?
                    .onStep { _, _ -> Observable<((), ())> in
                        XCTAssertEqual(0, workflow.completeCallCount)
                        return emptyObservable
                    }
                    .commit()
                return emptyObservable
            }
            .onStep { _, _ -> Observable<((), ())> in
                XCTAssertEqual(0, workflow.completeCallCount) // XCTAssertEqual failed: ("0") is not equal to ("1")
                return emptyObservable
            }
            .commit()
            .subscribe(())
        
        XCTAssertEqual(1, workflow.completeCallCount)
        XCTAssertEqual(1, workflow.forkCallCount)
        XCTAssertEqual(0, workflow.errorCallCount)
    }