ci: Add automated benchmark report

[samuelburnham]

• Updates manual benchmark workflow to run on a schedule, which runs all of the benchmarks every week
• Improves report units for proving and verifying time (60000 -> 1min0s)
• Upgrades to a custom Warpbuild r7iz.metal-32xl runner

Successful runs:

• https://github.com/argumentcomputer/zk-light-clients/actions/runs/11510874851/job/32043386696
• https://github.com/argumentcomputer/zk-light-clients/actions/runs/11510724999/job/32042984285

TODOs

• Reuse code between jobs
• Test thoroughly after merge

[tchataigner]

The fialure of the benchmark comes from the fact that the Sphinx logger was not enabled on Eth and KDA benchmarks. Activating the logger and running with log level info creates a 5 seconds overhead. I believe this is alright as the benchmarks will still give agood idea of the current performances, and we can run them without the logger for releases.

[huitseeker]

This is excellent, and I love the Zulip posts! Thank you so much for this.