[GraphBolt] high memory consumption when building GPU version

## 🔨Work Item

**IMPORTANT:**
* This template is only for dev team to track project progress. For feature request or bug report, please use the corresponding issue templates.
* DO NOT create a new work item if the purpose is to fix an existing issue or feature request. We will directly use the issue in the project tracker.

Project tracker: https://github.com/orgs/dmlc/projects/2

## Description
We're finding high memory consumption happens when building GPU version of DGL. This issue is firstly introduced when GPU sampling is added. Widely using `AT_DISPATCH_xxx` macros could be one of the culprit. And I think there's more culprits.
This issue make `conda-forge` struggling a bit. See details [here](https://github.com/conda-forge/dgl-feedstock/pull/28).

So we need to look into it.

## Depending work items or issues