buckets() flux query results truncated when using a non-administration token

__Steps to reproduce:__

1. Create a non-admin API token with "Read all buckets" permission.
2. Create more than 20 buckets whose name start with a letter.
3. Execute a Flux query containing a single `buckets()` statement using the HTTP API with the API token created at step 1.

__Expected behavior:__
Query results should contain all buckets created at step 2.

__Actual behavior:__
Query results only contain 18 buckets.

__Environment info:__

* InfluxDB versions against which the actual behaviour was seen:
* `InfluxDB 2.0.2 (git: 84496e507a) build_date: 2020-11-19T03:59:35Z`.
* `InfluxDB 2.1.1 (git: 657e1839de) build_date: 2021-11-09T03:03:48Z`
* System info: Not relevant.
* Other relevant environment details: None.

__Config:__
Nothing particular.

Reproduce the issue by creating a non-admin token with “Read all buckets” permission, creating more than 20 buckets, and running a single buckets() Flux query through the HTTP API. Compare the returned count with the expected bucket count and investigate the authorization or result-limit path; done means all matching buckets are returned for the non-admin token.

Written by the indexing model from the issue text.