Bug with subquery and JOIN operation within filters

A cube crashes when we use `field_2` as a filter but works well when we use `field_2` as dimensions.

**Steps to reproduce the behavior:**

1. Run a query with filters:

```json
{
"filters": [
{
"member": "model_b.field_2",
"operator": "set"
}
],
"dimensions": [
"model_b.field_2",
"model_b.cube_unique_id"
],
"order": {
"model_b.field_2": "asc"
}
}
```

and then receive an error:

```json
RangeMaximum call stack size exceeded
```

2. Run without filters:

```json
{
"dimensions": [
"model_b.field_2",
"model_b.cube_unique_id"
],
"order": {
"model_b.field_2": "asc"
}
}
```

**Minimally reproducible Cube Schema:**

```javascript
asyncModule(async () => {

cube('model_a', {
sql: `select * from table`,

dimensions: {
field_1: {
sql: `field_1`,
type: 'string'
},
},

});

cube('model_b', {
sql: `select * from table`,

dimensions: {

cube_unique_id: {
sql: `row_id`,
type: 'string',
primary_key: true
},

field_1: {
sql: `sum(${metric_1}) over ()`,
type: `number`,
sub_query: true,
propagate_filters_to_sub_query: true,
},

field_2: {
sql: `${model_a.field_1}`,
type: 'string'
},

},

measures: {
metric_1: {
sql: `metric_1`,
type: 'max'
}
},

joins: {
model_a: {
relationship: 'one_to_one',
sql: `${CUBE.field_1} = ${model_a.field_1}`
},

},

});
})

```

**Version:** 0.35.1

Use the minimally reproducible Cube schema as the entry point, then run the filtered and unfiltered JSON queries to compare their behavior. Trace the interaction between the sub_query, propagate_filters_to_sub_query, and model_a join; done means the filtered query no longer raises a call-stack error and returns successfully.

Written by the indexing model from the issue text.