Package load time regressions in 1.12

[KristofferC]

I think this deserves a separate issue to track this.

❯ julia +nightly --project=@MTK_1.13 -e '@time using ModelingToolkit'
  3.292820 seconds (8.19 M allocations: 478.546 MiB, 7.74% gc time, 4.13% compilation time: 36% of which was recompilation)

  ❯ julia +1.11 --project=@MTK_1.11 -e '@time using ModelingToolkit'
  2.454841 seconds (5.07 M allocations: 346.203 MiB, 7.39% gc time, 3.76% compilation time: 31% of which was recompilation)

  ❯ julia +1.10 --project=@MTK_1.10 -e '@time using ModelingToolkit'
  2.507505 seconds (4.80 M allocations: 359.739 MiB, 10.68% gc time, 2.24% compilation time: 34% of which was recompilation)

Nightly has a big bump in load time.

[KristofferC]

This seems to be worse on 1.12-nightly?

julia> VERSION
v"1.12.0-DEV.2136"

julia> @time using ModelingToolkit
  3.713581 seconds (9.52 M allocations: 519.955 MiB, 8.09% gc time, 2.98% compilation time: 31% of which was recompilation)

[ufechner7]

Julia 1.12-beta2 is still looking very bad:

hyperfine 'julia +1.10 --startup-file=no -e "using KiteModels"' 'julia +1.12 --startup-file=no -e "using KiteModels"'
Benchmark 1: julia +1.10 --startup-file=no -e "using KiteModels"
  Time (mean ± σ):      5.558 s ±  0.088 s    [User: 5.205 s, System: 0.922 s]
  Range (min … max):    5.478 s …  5.746 s    10 runs
 
Benchmark 2: julia +1.12 --startup-file=no -e "using KiteModels"
  Time (mean ± σ):     10.003 s ±  0.050 s    [User: 10.030 s, System: 0.548 s]
  Range (min … max):    9.916 s … 10.074 s    10 runs
 
Summary
  julia +1.10 --startup-file=no -e "using KiteModels" ran
    1.80 ± 0.03 times faster than julia +1.12 --startup-file=no -e "using KiteModels"