Report host stats on TPCH benchmarks

[gabotechs]

Reports some relevant data about the environment in which the TPCH benchmarks are running:

==== Comparison with the previous benchmark from 2025-09-08 13:28:37 UTC ====
os:        macos
arch:      aarch64
cpu cores: 16
threads:   2 -> 4
workers:   3 -> 6
=============================================================================
 Query 1: prev= 231 ms, new= 120 ms, diff=1.93 faster ✅
 Query 2: prev= 159 ms, new= 142 ms, diff=1.12 faster ✔
 Query 3: prev=  58 ms, new=  52 ms, diff=1.12 faster ✔
 Query 4: prev= 114 ms, new=  60 ms, diff=1.90 faster ✅
 Query 5: prev= 297 ms, new= 183 ms, diff=1.62 faster ✅
 Query 6: prev=  26 ms, new=  14 ms, diff=1.86 faster ✅
 Query 7: prev= 298 ms, new= 194 ms, diff=1.54 faster ✅
 Query 8: prev= 136 ms, new= 122 ms, diff=1.11 faster ✔
 Query 9: prev= 203 ms, new= 149 ms, diff=1.36 faster ✅
Query 10: prev= 399 ms, new= 313 ms, diff=1.27 faster ✅
Query 11: prev=  72 ms, new=  59 ms, diff=1.22 faster ✅
Query 12: prev= 134 ms, new=  88 ms, diff=1.52 faster ✅
Query 13: prev=  99 ms, new=  70 ms, diff=1.41 faster ✅
Query 14: prev=  77 ms, new=  45 ms, diff=1.71 faster ✅
Query 15: prev=  71 ms, new=  72 ms, diff=1.01 slower ✖
Query 16: prev=  87 ms, new=  57 ms, diff=1.53 faster ✅
Query 17: prev= 211 ms, new= 141 ms, diff=1.50 faster ✅
Query 18: prev= 586 ms, new= 297 ms, diff=1.97 faster ✅
Query 19: prev= 111 ms, new= 101 ms, diff=1.10 faster ✔
Query 20: prev=  92 ms, new=  66 ms, diff=1.39 faster ✅
Query 21: prev= 409 ms, new= 170 ms, diff=2.41 faster ✅
Query 22: prev=  57 ms, new=  34 ms, diff=1.68 faster ✅

[jayshrivastava]

Nice!

[NGA-TRAN]

Th ePR looks good. Some questions regarding cores & threads

[NGA-TRAN]

Is this the number of cores per worker or total of cpu cores of all workers?
Is there any constraints between this and number of threads per worker? Should number of threads per worker < CPU cores per worker? Is there any enforcement or unexpected behavior if it is not enforced

[gabotechs]

Is this the number of cores per worker or total of cpu cores of all workers?

Total CPU cores available in the machine that runs the benchmarks

Is there any constraints between this and number of threads per worker?

Only the ones specified by the people running the commands, it's possible to do --threads 1000 in a 2 core CPU machine. It does not make much sense to do that though...

Should number of threads per worker < CPU cores per worker?

Assuming get_available_parallelism() returns the number of CPU cores in the machine, I think so yes. As these are just benchmarks meant to be run by us, I think it's fine to not do any enforcement in the code, hopefully people running the benchmarks know what they are doing if the do --threads 1000 in a 2 core CPU machine.

[NGA-TRAN]

I agree people who run the benchmark should know what we are doing. Maybe we just help remind them that. How about we add a brief message saying number of threads should be < number of cores and number of partitions should be < number of threads?

[gabotechs]

Added a doc comment in the CLI