Add a real benchmark-alteration test #480

The goal is to introduce an actual performance change, rather than editing benchmark output.

performance change could rely on:

• debug versus release builds
• artificial memory accesses
• compiler flags

(see parent issue)
Verify that the comparison script detects the expected FAST or SLOW outcomes.