Deeply Nested Aggregate query is not filtered by parent colum

The query needs to count each occurance of a column value in a "history" table per each "definition" row. This requires looking up possible values and counting each of the possible "ExitReason Codes".

This query gets each **subtask**, but the **count** is the total count of each ExitReason, and not the ExitReason count for the give subtask:

```gql
query SubtaskExitReasonCount {
ivr_subtasks {
friendly_name
subtask_histories {
exitReasonByExitReason {
exit_reason
subtask_histories_aggregate {
aggregate { count }
}
}
}
}
}
```

The following does work as expected; however, it would require filtering by each `subtask_id` rather depending on the implied relationships.

```gql
query SubtaskExitReasonCountWithFilter {
ivr_subtasks {
friendly_name
subtask_histories {
exitReasonByExitReason {
exit_reason
subtask_histories_aggregate(where: {subtask_id: {_eq: 1}}) {
aggregate { count }
}
}
}
}
}
```

- How can the `subtask_histories_aggregate` be filtered/grouped by the `subtask_id`?
- Must this be done with a `VIEW` with an extra group by statement?
- Can a where clause reference a parent relationship property?

```gql
## Pseudo Code
subtask_histories_aggregate(where: {subtask_id: {_eq: "_root.id" }}) {
aggregate { count }
}
```

Start by reproducing the nested aggregate behavior with the two GraphQL queries shown in the issue, then trace how the relationship and aggregate query are generated. Done means the aggregate count is scoped to each parent subtask without requiring an explicit subtask_id filter, with regression coverage for the provided case.

Written by the indexing model from the issue text.