Render the segmentation as mesh

[manasaV3]

Impact: Render the segmentation layers as 3d surface mesh to make them more performant.

The current strategy for rendering the segmentation in neuroglancer is to render it as a image volume, as it is a binary mask, we will be able to optimize it by rendering it as a surface mesh.

Note: There are some known issues with neuroglancer rendering multiscale surface mesh. We should precompute a multiscale mesh to handle the issues.

[manasaV3]

The support for single-resolution mesh in neuroglancer is a legacy feature.
We would have to navigate some known pitfalls for the multiresolution mesh. There are code snippets available, that will help us with that, but might require a little poking around.