Inability to manipulate range offsets in dashboards

InfluxDB 2.0.2

Original task is simple I want to have on a plot some data for closen interval and data for previous interval with fixed shift/offset on the same plot (i.e shifted data will be shifted to current interval)
Shift is solved but timeShift() , but selection is not solvable.

So, let assume typical query:
```
from(bucket: "bucket")
|> range(start: v.timeRangeStart, stop: v.timeRangeStop)
|> filter(fn: (r) => r["_measurement"] == "hmeasurement")
|> filter(fn: (r) => r["_field"] == "field")
|> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)
|> yield(name: "mean")
```

Problem is to have another query in a equivalent form of
```
from(bucket: "bucket")
|> range(start: v.timeRangeStart-24h, stop: v.timeRangeStop-24h)
|> timeShift(duration: 24h)
|> filter(fn: (r) => r["_measurement"] == "hmeasurement")
|> filter(fn: (r) => r["_field"] == "field")
|> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)
|> yield(name: "mean")
```

And in this query, particularly, line
```
|> range(start: v.timeRangeStart-24h, stop: v.timeRangeStop-24h)
```
is not solvable.
you can use custom variable for ```start:``` here, but using anything but ```v.timeRangeStop``` for the ```stop:``` (or custom variable which is defined like ```t=v.timeRangeStop```) , leads to the error ```record is missing label windowPeriod```. (I tried to alter ```t``` with add/subDuration)

So, unless there is something obvious I am missing, there is a clear bug or missing feature. And I don't think it is such exotic use case. (let say you want to compare this week performance to last week performance).

Maybe, I am wrong, but I presume this should work:
```
t0=experimental.subDuration(from: v.timeRangeStart,d: 24h)
t1=experimental.subDuration(from: v.timeRangeStop,d: 24h)

from(bucket: "bucket")
|> range(start: t1, stop: t2)
|> timeShift(duration: 24h)
|> filter(fn: (r) => r["_measurement"] == "hmeasurement")
|> filter(fn: (r) => r["_field"] == "field")
|> aggregateWindow(every: v.windowPeriod, fn: mean, createEmpty: false)
|> yield(name: "mean")
```

data-wise, this query is all right.
And I can successfully query in CLI:
```
import "experimental"

t1=experimental.subDuration(from: now(), d: 24h)
t0=experimental.subDuration(from: t1,d: 24h)

from(bucket: "bucket")
|> range(start: t1, stop: t2)
|> timeShift(duration: 24h)
|> filter(fn: (r) => r["_measurement"] == "hmeasurement")
|> filter(fn: (r) => r["_field"] == "field")
|> aggregateWindow(every: 15m, fn: mean, createEmpty: false)
|> yield(name: "mean")
```

But using the very same query in query editor in InfluxDB dashboard UI results in ```record is missing label windowPeriod``` error.

Reproduce the supplied query in the dashboard query editor and compare it with the CLI result. Focus on the dashboard UI's handling of shifted range expressions and the `windowPeriod` error; done means the equivalent previous-interval query runs in the dashboard without that error.

Written by the indexing model from the issue text.