Clean up icount benchmarks

* Delete some memcpy tests that were a bit excessive
* Always use the same offset of 65
* Add a memmove test with aligned source and destination
* Improve printing output and add more comments
* Use a constant for 1 MiB so it shows up in the benchmark logs

Author: tgross35