Small vectors in various sizes. These store a certain number of elements inline, and fall back to the heap for larger allocations. This can be a useful optimization for improving cache locality and reducing allocator traffic for workloads that fit within the inline buffer.