Enablement work for future core shader support

[embeddedt]

This PR lays some early groundwork for a hybridized renderer that is capable of using vanilla core shaders (if necessary) and more efficient shaders otherwise.

These changes are not sufficient to use resource packs that modify terrain shaders. If you are a user, do not expect that builds from this PR will function with core shaders.

The change set is as follows:

• Introduce VanillaShaderChunkRenderer, which draws chunks from our vertex buffers using the appropriate vanilla core shader. This renderer is rather inefficient compared to the regular one, as it only issues multidraw calls for one section at a time, rather than one region. This is because the vanilla shader expects the chunk offset uniform to be updated for each section.
• Introduce a third vertex type (used internally) which produces data in the same format as DefaultVertexFormat.BLOCK.
• Clean up vertex types so that they are selected in one central location.
• Add a system property for testing VanillaShaderChunkRenderer (otherwise, the normal renderer is used).

More work is needed for this to be usable in practice. In particular:

• It needs to be possible to select separate vertex types for each render pass. Otherwise, all chunks must use the vanilla-like vertex type, which is problematic both for performance and for the implementation (as there isn't space for the extra data our shader needs from the vertex).
• Some simple core shaders need to be found to do testing with.
• Currently, the Embeddium renderer only has three render passes (solid/cutout/translucent); data from the other passes is merged. This will likely need to be disabled if resource packs override the tripwire or mipped cutout shaders.
• Sugarcane disappears at a significant enough distance when testing with the vanilla terrain shader. This is likely related to the point above.