[performance][medium] Reduce allocations and string copies when generating command lists

[upupwrite]

Summary

Interactive flows such as generateList()/interactiveSearch build strings and data structures each time a search is performed. For large numbers of commands or directories, string allocations and copying may dominate runtime and cause latency. The UI builds concatenated strings for display each time.

Reproduction / evidence

• src/search.cpp: generateList constructs a long string by iterating all paths and commands. Each invocation will allocate memory proportional to the total size.

Impact

• Increased latency for generating lists and for interactive sessions with large datasets.
• Higher memory churn leads to garbage in runtime and CPU overhead.

Suggested fixes / improvements

1. Use more efficient data structures for display generation:

   • Avoid building a single giant std::string; use streaming to stdout or buffered writers that pre-allocate capacity when the size is known.
   • Reuse buffers across calls when possible.

2. Implement pagination / lazy evaluation:

   • Only generate and render portions of the list visible to the user (or requested), especially when integrated with fzf or pagers.

3. Profile hot paths and optimize critical loops:

   • Use tools like perf or valgrind to find hotspots and inline small helpers, minimize copies, and use references where appropriate.

Files/places to review first

• src/search.cpp (generateList, interactiveSearch)

Severity: medium