[Feature Request] Non-uniform grid resolution support

I'm doing experiements with the 3-dim hash grid based input encoding, which used the normalized position inside a scene as the position input. However, some artifacts become visually obvious when the scene becomes long and narrow, more specifically, when the range of one axis of the bounds is significantly larger than the rest two. Several efforts were made to suppress these artifacts, but I think a more straightforward solution would be adding the support for non-uniform grid resolution in different axis. Current implementation considers the same base_resolution for all dimensions, it would be great if it takes a list as a input such that we can specify different base_resolution configuration for different dimensions like `"base_resolution": [4, 4, 16]`. I wonder if it is possible to make some modifications to achieve this? Appreciate your efforts on this wonderful project!