Better docs/name needed for "multi-threaded" benchmarking

Would it have been better if any of this would be in docs / is there anything brought up in this issue appears to be omitted in docs?

I think so, yes. Perhaps a sentence or two on the README regarding how the "multi-threaded" implementation spawns self-contained benchmark instances, and users should expect memory footprint for the benchmark program to grow linearly in the number of threads.

Originally posted by @jmftrindade in https://github.com/google/benchmark/issues/806#issuecomment-488472617