There are two main optimizations:
1. We have optimized Module.concat/2 (it is roughly 100% faster)
2. We have optimized rendering of Inspect.Algebra by skipping
the intermediate s_doc representation. This reduces memory
allocation and skips one extra pass on rendering.