polkadot-omni-node: Fully remove benchmark sub-command

[kianenigma]

... And make frame-omni-bencher work for all use cases, such as: #5303

Original issue for @serban300:

Remove the benchmark commands from polkadot-parachain: see comments 1, 2, 3: Here we need to wait for the CI bot bench commands to be migrated to omni-node first:

• Moved presets to the testnet runtimes #5327
• [benchmarking] Reset to genesis storage after each run #5655
• Enable GHA short benchmarks with frame-omni-bencher #5706
• Short benchmarking in CI with frame-omni-bencher + extract chain-spec-builder stuff to get_preset polkadot-fellows/runtimes#379

[kianenigma]

Blockers towards this are:

• CI short benchmarking with frame-omni-bencher  #4405
• frame-omni-bencher fix for --repeat > 1 #5083
• /cmd bot followups #5545

[kianenigma]

@bkontur @mordamax it seems like you are taking steps towards making the omni bencher work. Can you please share how you feel about commitment and the amount of work that is left? if you need any support?

As you see in the project board, this is not super time sensitive for omni-node, but it is good to get an idea.

[mordamax]

re #5545 i hope should be done and merged within this week

[bkontur]

From my perspective, this is also a blocker or prerequisite: #5327 (refactoring testnet runtimes to use get_preset from the polkadot-parachain chain-specs; a few more runtimes need to be refactored). I could possibly finish it next week, or another volunteer could take it on.

After that, we need to fix frame-omni-bencher #5083. We have a hotfix/workaround here, but I think there's a bug "elsewhere," and I'd like to investigate the real cause before we merge the hotfix/workaround. I plan to look into it next week, as I'm currently focused on higher-priority tasks.

Once both are fixed/merged, we can proceed with merging #4405, which enables short benchmarks with frame-omni-bencher for CI.

After that, we can safely remove the benchmarking code from omni-node.

[skunert]

I think the latest development in this discussion was that omni-node should still expose benchmarking commands, but they should call into the same code also used by frame-omni-bencher.