Bug: Dynamic query zoom breaks when using dashboard variables

Nobody has claimed this yet.

## About the bug

**Steps to reproduce:**
List the minimal actions needed to reproduce the behavior.

1. Create a dashboard with (a) a dashboard variable, and (b) a graph using that variable in its query.
2. Zoom into the graph to trigger a re-query of the data at finer granularity

**Expected behavior:**

The graph displays more granular data from the new query.

**Actual behavior:**

- The console shows an HTTP 400 error which suggests the variable is not being included in the API request for new data; and
- The graph fails to show more granular data since it does not have new query results (only old coarse ones from the original query)

**Visual Proof:**

Here's a dashboard graph of some query results which depend on a dashboard variable, with the time range set to the 7 days:

Here's what I see if I click-and-drag to zoom into the first ~5.5 hours of the graph; note the lack of time resolution in the data:

In contrast, here's what I see if I manually update the time range (via the dropdown at the top right of the dashboard UI) to the same 5.5 hour window instead of using the zoom feature:

The browser console displays a 400 error for a POST request to the `/api/v2/query?orgID=...` endpoint initiated by `query.ts:61`, and the response payload for that request contains the following JSON:

```json
{
"code": "invalid",
"message": "error @5:40-5:41: record is missing label [redacted_variable_name]"
}
```

## About your environment

**Environment info:**

Cloud version of Influx 2 (TSM), so no shell access to run the requested commands.

Start at query.ts:61, where the zoom-triggered POST request to /api/v2/query is initiated, and reproduce the dashboard-variable case described in the issue. Done means the variable is included in the zoom re-query, the request no longer returns HTTP 400, and the graph displays more granular data.

Written by the indexing model from the issue text.