Clarifying resource group token distribution behavior

I'm currently in the process of root causing why our TiDB host fleet sees RG exceeded errors for resource groups with < 50% avg utilization / < 75% max utilization. I'm looking to clarify [this](https://github.com/tikv/pd/blob/640fbed566acac2e26db2b832b68c8d41a35d1e1/client/resource_group/controller/controller.go#L1257) part of the token allocation logic:

```
// Otherwise the granted token is delivered to the client by fill rate.
cfg.NewTokens = 0
trickleDuration := time.Duration(trickleTimeMs) * time.Millisecond
deadline := gc.run.now.Add(trickleDuration)
cfg.NewRate = float64(bucket.GetSettings().FillRate) + granted/trickleDuration.Seconds()
```

AFAICT, we should set NewTokens to the granted value here as well, instead of updating the FillRate. Can someone confirm the code really reflects the intended behavior?