Pre-aggregations doesn't work well with sub-queries

[saurabh0402]

Describe the bug
I have a measure that uses a dimension with sub_query and propagate_filters_to_sub_query set to true. When creating a pre-aggregation in the same cube and querying the measure, it errors out with the following error message

RequestError: Error: Internal: Error during planning: Table or CTE with name '<table name>' not found

To Reproduce
I used the following schema to create the cube

cubes:
  - name: forecasts
    sql: select * from schema.forecasts
    pre_aggregations:
      - name: first
        measures:
          - max_cutoff
        dimensions:
          - product
          - cutoff
        external: true
    measures:
      - name: max_cutoff
        sql: cutoff
        type: max
      - name: mean_absolute_percentage
        sql: abs((y - yhat) / greatest(y, 1e-16))
        type: avg
        filters:
          - sql: "{CUBE}.cutoff >= {tmp_max_cutoff} - INTERVAL '12 Weeks'"
    dimensions:
      - name: product
        sql: product_id
        type: string
        primary_key: true
        shown: true
      - name: cutoff
        sql: cutoff
        type: time
        primary_key: false
      - name: tmp_max_cutoff
        sql: "{max_cutoff}"
        type: number
        sub_query: true
        propagate_filters_to_sub_query: true
      - name: date
        sql: date
        type: time

I am then running the following query

const resultSet = await cubeApi.load({
    measures: ['forecasts.mean_absolute_percentage'],
    dimensions: ['forecasts.product'],
});

This error out with the message given above.
I think this is happening because, Cube is detecting that it can be satisfied via the pre-aggregation and thus runs query on pre-aggregation only. This causes the sub-query to fail.

Expected behavior
Cube should detect the sub-query and not run this specific query on pre-aggregations but use the source database.

Version:
0.36.1

Additional Comments

• One workaround is to create two different cubes one with the pre-aggregation and one without it and add the measure with sub-query to the one without pre-aggregation. If there's a better workaround, please let me know.