[1.21.1] Feature: Add cached ber rendering pipeline

[ZhuRuoLing]

"cached" means the rendering pipeline will cache those verticles put by a custom renderer.

To use the rendering pipeline, users would implement the method IBlockEntityRendererExtension#renderCached , then notify the pipeline that there has something changed and the vertex cache need to be rebuilt.

The entire pipeline looks like some sort of chunk geometry, but users do not need to register the RenderTypes they used into a list.

This rendering pipeline can provide exceptional performance when rendering fixed vertices for BER and has great compatibility with other optimization mods like Embeddium and Sodium.

[CLAassistant]

All committers have signed the CLA.

[neoforged-pr-publishing]

• Publish PR to GitHub Packages

[Technici4n]

This is an interesting PR. We need to be careful not to accept something that will bring a large maintenance burden though.

Out of curiosity, how does this PR compare to static chunk geometry?

[ZhuRuoLing]

This PR supports custom RenderTypes without registering them into RenderType#CHUNK_BUFFER_LAYERS, and using a custom RenderType in static chunk geometry might be very hard to implement under optimization mods like Sodium.

Besides these, it doesn't differ much from static chunk geometry, regardless of whether optimization mods are installed.

[neoforged-automation]

@ZhuRuoLing, this pull request has conflicts, please resolve them for this PR to move forward.