Decision Tables in Multi-Instance Subprocesses Write Output to Main Process

**Describe the bug**
When you attempt to use a decision task inside of a multi-instance subprocess that operates in parallel, the output variables get written to the main process. This causes all instances of the multi-instance subprocess to "see" the outputs from the last executed decision table.

**Expected behavior**
Output variables from decision tasks should be scoped in such a way to be usable inside of parallel multi-instance subprocesses

**Code**
Sample app attached. The sample app contains a groovy script task that creates an `items` array containing the numbers 1, 2, & 3. Following that there is a parallel multi instance sub-process containing a decision task and a user task with a form. Decision task takes the `item` passed into the specific instance and uses one a rule to write it back out to `item_out`. The user tasks form supplies a way to view `item`, `item_out`, and a `p_var` process variable set by the script task. Looking at the executions of the decision table in the flowable-admin app, it's clear that each instance is operating on the correct `item` variable and putting out a correct output, but only the last `item_out` is displayed for all forms.

Beginne mit der angehängten Beispiel-App und reproduziere den parallelen Multi-Instance-Subprozess, der die Decision Task und die User Task enthält. Untersuche, wie die Decision Execution in item_out schreibt und wie die Formulare item, item_out und p_var lesen; fertig ist die Aufgabe, wenn jede Instanz ihre eigene Decision-Ausgabe anzeigt statt des Werts der zuletzt abgeschlossenen Instanz.

Vom Indexierungsmodell aus dem Issue-Text verfasst.