Tests fail on 1.11

[KristofferC]

┌ Info: Running IR test in a subprocess...
│   cmd = `/Users/kristoffercarlsson/julia1.11/usr/bin/julia -C native -J/Users/kristoffercarlsson/julia1.11/usr/lib/julia/sys.dylib --depwarn=yes --check-bounds=yes -g1 --color=yes --startup-file=no --startup-file=no --color=yes --check-bounds=no --code-coverage=none --inline=yes`
└   script = "/Users/kristoffercarlsson/JuliaPkgs/Setfield.jl/test/perf.jl"
heads_lens = [:call, :call, :call, :call, :call, :call, :call, :call, :call, :new]
heads_hand = [:call, :call, :call, :new]
IR: Test Failed at /Users/kristoffercarlsson/JuliaPkgs/Setfield.jl/test/perf.jl:83
  Expression: uniquecounts(heads_lens) == uniquecounts(heads_hand)
   Evaluated: Dict(:call => 9, :new => 1) == Dict(:call => 3, :new => 1)

A comment in the test says that this test might be too strict. But maybe there is a regression in some optimization that used to happen here.

[jw3126]

probably makes sense to manually inspect @btime and @code_*