Add run-perf.sh

[vdukhovni]

This is just the run-perf.sh script for sharing with the team, work-in-progress, do not merge.

[Sashan]

I think this is good enough. I gave it try. to speed things up I've modified some parameters:

• reduced number of iterations to two
• reduced number of threads to 4 and 8
• reduced versions to 1.1.1, 3.0, 3.3 and master

works flawlessly. I think it's good enough. I would not waste too much time with it. What we really need to decide are parmeters which we will be using for testing on all platforms:

• number of iterations: the first version from @nhorman got 50, I've reduced the number to 25, but it still feels too much
• versions I think @mattcaswell proposal to test 1.1.1, 3.0, 3.3 and master is sufficient
• number of threads I like what @nhorman suggested: 1 2 4 8 16 32 64 128

with parameters like above we can expect test duration on ryzen 7 with 8 cores roughly 12 hours.

[vdukhovni]

Applied a bit more polish and trimmed the version list. Left iterations and threads alone for now, pending future decisions. Should be good to merge.

[nhorman]

merged