Optimize concatenation of complex data type, such as list, struct

I've found that the concatenation of complex data types will be the bottleneck when I worked on an internal performance issue.

The issue names no files, tests, or entry points and provides only a screenshot of a performance bottleneck. Start by locating the concatenation implementation for complex data types and reproducing the reported workload with a benchmark. Done means demonstrating an improvement without changing concatenation behavior.

Written by the indexing model from the issue text.