Add profiling guide to help identify performance problems

[rdw-software]

Should compare multiple approaches to profiling:

• Using os.time() and similar naive facilities
• Using time or perf (Unix only) and other OS-level facilities
• Using the LuaJIT profiler (docs here)
• Using C++/C profilers, via C and FFI API? (e.g. ShinyProfiler - what is the industry standard in those spaces?)

See: Flame graphs? (https://www.brendangregg.com/FlameGraphs/cpuflamegraphs.html)

Also needs to make sure people understand the risks of premature optimization, and when performance actually matters. Plus, knowing to benchmark vs. "guessing".