Single multi groupby v2

[jayzhan211]

Which issue does this PR close?

Closes #.

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

[jayzhan211]

// this change
// Gnuplot not found, using plotters backend
// Benchmarking benchmark: Warming up for 3.0000 s
// Warning: Unable to complete 10 samples in 5.0s. You may wish to increase target time to 8.2s or enable flat sampling.
// benchmark               time:   [150.22 ms 154.31 ms 158.06 ms]
//                         change: [+254.76% +261.92% +270.28%] (p = 0.00 < 0.05)
//                         Performance has regressed.
// Found 1 outliers among 10 measurements (10.00%)

// main
// Gnuplot not found, using plotters backend
// Benchmarking benchmark: Warming up for 3.0000 s
// Warning: Unable to complete 10 samples in 5.0s. You may wish to increase target time to 29.0s.
// benchmark               time:   [2.0650 s 3.5149 s 5.1064 s]
//                         change: [+1182.7% +2229.6% +3244.4%] (p = 0.00 < 0.05)
//                         Performance has regressed.

[jayzhan211]

I got 2.1296 for main

[jayzhan211]

single-multi-groupby-v2

Gnuplot not found, using plotters backend
// benchmark               time:   [41.512 ms 41.931 ms 42.318 ms]
//                         change: [-6.5657% -5.6393% -4.6452%] (p = 0.00 < 0.05)
//                         Performance has improved.
// Found 1 outliers among 10 measurements (10.00%)
//   1 (10.00%) high mild

main

Gnuplot not found, using plotters backend
benchmark               time:   [43.127 ms 43.299 ms 43.636 ms]
                        change: [-99.150% -98.764% -97.880%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high severe

[jayzhan211]

I found removing pre-partition cause huge regression for many queries