feat(runtime, ci): finish migration to bare-metal benchmark runner and update weights

[ByteNacked]

What was done:

A migration of the benchmark runner was performed, moving from a virtual host to a bare-metal host to increase the reproducibility of benchmark results.

Since the bare-metal host differs somewhat in performance from the equivalent virtual host, we investigated these differences through several benchmark runs and calculated the average discrepancy between them. As a result, a 10% adjustment was introduced to the instruction weights, aligning the instruction weight results with those on the virtual host.

Additionally, for the multi-threaded benchmarks of the gear_builtin pallet, it was decided to fix the execution of these benchmarks to 4 CPU cores, again to improve the reproducibility of the results.

• Added script which adjust Instruction weights by 10 % when running on bare metal runner
• Run multi threaded benches (pallet_gear_builtin) on fixed number of cores (4 cores)
• Update weights from bare-metal run

@reviewer-or-team

[ByteNacked]

Should InstructionWeight version be incremented in this PR due to some instruction weight change?

[breathx]

Should InstructionWeight version be incremented in this PR due to some instruction weight change?

Unnecessary, it's all up to you, on release tagging I'll anyway double check all versions

[breathx]

Please resolve master brach conflicts