evmone: upgrade to version 0.13.0

[chfast]

Upgrade evmone to version 0.13.0 by rebasing erigontech's fork of it: https://github.com/erigontech/evmone/tree/rebase_to_v0.13.
Upgrade intx to version 0.12.0.

The new version of evmone manages ExecutionState internally and
removes it from public API. This makes object pools for
ExecutionStates in Silkworm unnecessary.

[yperbasis]

It looks like EOF will be postponed until PectrB (=Osaka?)

[chfast]

This reflects the status of evmone, but actually we can put false here for now.

[chfast]

Set to false with a comment.

[chfast]

Please don't merge yet. I want to make official evmone release first.

[chfast]

This reflects the status of evmone, but actually we can put false here for now.

[chfast]

Mainnet execution benchmarks

We run from scratch Mainnet blocks execution and check how quickly it reaches block 15M (no disk space for more).

• base (1695 mins)

  INFO [09-27|17:41:08.674 UTC] Resource usage                           mem=55.43 GB chain=1.02 TB temp=0.00 B uptime=1d 4h 15m
  INFO [09-27|17:41:09.036 UTC] [5/12 Execution]                         block=14996203 blocks/s=49 txns/s=8763 Mgas/s=765
  INFO [09-27|17:41:39.036 UTC] [5/12 Execution]                         block=14997745 blocks/s=51 txns/s=9302 Mgas/s=791
  INFO [09-27|17:42:09.036 UTC] [5/12 Execution]                         block=14999449 blocks/s=56 txns/s=9912 Mgas/s=872
  INFO [09-27|17:42:39.037 UTC] [5/12 Execution]                         block=15001089 blocks/s=54 txns/s=10103 Mgas/s=844
  

• this PR with execution states in thread_local (1750 mins, +3%)
  This run is slower by 3%. It also differences from the PR by hacking evmone to place the execution storage array in thread_local storage instead of having it attached to evmone::VM. The reason for the hack is Extend evmone::VM lifetime #2383. However, it looks like this change the performance (before the hack the performance was the same ~3% slower).

  INFO [09-29|20:52:19.536 UTC] Resource usage                           mem=55.41 GB chain=1.02 TB temp=0.00 B uptime=1d 5h 10m                                                                                                                                                 
  INFO [09-29|20:52:19.922 UTC] [5/12 Execution]                         block=15001155 blocks/s=53 txns/s=10078 Mgas/s=831                                                                                                                                                      
  

• base 2nd run (1715 mins, +1%)

  INFO [10-01|13:40:26.474 UTC] Resource usage                           mem=55.49 GB chain=1.02 TB temp=0.00 B uptime=1d 4h 35m                                        
  INFO [10-01|13:40:26.791 UTC] [5/12 Execution]                         block=14992723 blocks/s=45 txns/s=8157 Mgas/s=704                                              
  INFO [10-01|13:40:56.791 UTC] [5/12 Execution]                         block=14994328 blocks/s=53 txns/s=9439 Mgas/s=822                                              
  INFO [10-01|13:41:26.791 UTC] [5/12 Execution]                         block=14995830 blocks/s=50 txns/s=8804 Mgas/s=773                                              
  INFO [10-01|13:41:56.791 UTC] [5/12 Execution]                         block=14997304 blocks/s=49 txns/s=8827 Mgas/s=757                                              
  INFO [10-01|13:42:26.792 UTC] [5/12 Execution]                         block=14998928 blocks/s=54 txns/s=9632 Mgas/s=833                                              
  INFO [10-01|13:42:56.792 UTC] [5/12 Execution]                         block=15000567 blocks/s=54 txns/s=10003 Mgas/s=840                                             
  

• this PR with execution states in thread_local and intx perf regression fixed (1715 mins, +1%)

  INFO [10-03|17:24:27.348 UTC] Resource usage                           mem=55.47 GB chain=1.02 TB temp=0.00 B uptime=1d 4h 35m                                                                                                                                                 
  INFO [10-03|17:24:27.640 UTC] [5/12 Execution]                         block=15001546 blocks/s=55 txns/s=10585 Mgas/s=856                                                                                                                                                      
  

[canepat]

Please don't merge yet. I want to make official evmone release first.

I'm converting this PR to draft again just to avoid accidental merge.

[chfast]

I've discovered intx 0.12 performance regression at least with GCC 13. The fix is this: chfast/intx#325, but I need to investigate what exactly is going on. I'm running another Mainnet benchmark and will report back when finished, but the initial number are looking good.

[JacekGlen]

Apart from intx performance, the rest looks good

[chfast]

I reported the issue related to GCC-13 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117000.
In GCC-14 the thing is gone, and intx 0.13 works faster than intx 0.12. But also slower than GCC 13 :) I will be tracking this but we can't fix everything at once.