Add benchmark workflow (RPi4 / Arm Cortex-A72)

[potsrevennil]

This PR depends on #68

I Added a benchmark workflow. Note that the workflow will only run on main branch or PR labeled as benchmark. The tests script is also refactored a bit in order to show the benchmark summary on GitHub action.

I refactored the nix shell specification in order to speed up ci, and also
disabled ci jobs running on self-host runner when pushed to fork repositories by the way.

[hanno-becker]

I'd like to make the --cpu to tests bench optional and omit -march if -mcpu is specified.

[hanno-becker]

Thank you @potsrevennil, I confirm that --arch-flags works.

May I ask for one more usability improvement: For perf, I have to run the binaries with sudo, but somehow sudo ./scripts/test ... doesn't work (and is also not ideal, because the build itself really needs not sudo rights).

Can you add an option to the test script which builds only, and prints the names of the files that have been produced?

[potsrevennil]

@hanno-becker Yeah, I agree with you that it indeed makes more sense in some cases to separate building and running the binary. There're also cases that we would like to run the binaries in a different machine.

However, I'm thinking this might a bit out of scope for this PR, maybe we could open an issue to address this more, and make it a separate PR. What do you think ?

[hanno-becker]

@potsrevennil Ok!