Module Tolk.Codegen_lower

Codegen lowering after optimization.

lower runs expansion, devectorization, GPU dimension mapping, image lowering, index dtype concretization, decompositions, and renderer-specific rewrites.

This module is the post-optimization half of tinygrad's codegen/__init__.py pipeline. It intentionally has no dependency on Search, Postrange, or Heuristic, so beam search can lower and linearize candidate schedules without depending on Codegen.

lower renderer sink is sink after the non-optimization codegen passes. The result is ready for Linearizer.linearize.