Even faster NaN checks

[github-actions[bot]]

Pull Request Test Coverage Report for Build 4782969121

• 10 of 10 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.07%) to 89.44%

---

Totals
Change from base Build 4774460028:	-0.07%
Covered Lines:	991
Relevant Lines:	1108

---

💛 - Coveralls

[github-actions[bot]]

Benchmark Results

	master	275eae805ee7c4...	t[master]/t[275eae805ee7c4...]
OperatorEnum/ComplexF32/evaluation	0.0363 ± 0.0012 ms	23.5 ± 2.1 μs	1.54
OperatorEnum/ComplexF64/evaluation	0.0387 ± 0.0028 ms	0.0362 ± 0.0026 ms	1.07
OperatorEnum/Float32/derivative	0.137 ± 0.0023 ms	0.134 ± 0.0022 ms	1.03
OperatorEnum/Float32/derivative_turbo	0.166 ± 0.014 ms	0.157 ± 0.002 ms	1.06
OperatorEnum/Float32/evaluation	8.25 ± 0.53 μs	8.48 ± 0.56 μs	0.973
OperatorEnum/Float32/evaluation_turbo	7.34 ± 0.66 μs	7.48 ± 0.76 μs	0.981
OperatorEnum/Float64/derivative	0.203 ± 0.005 ms	0.186 ± 0.0042 ms	1.09
OperatorEnum/Float64/derivative_turbo	0.226 ± 0.0077 ms	0.205 ± 0.0039 ms	1.1
OperatorEnum/Float64/evaluation	12.3 ± 1.2 μs	12.5 ± 1.3 μs	0.984
OperatorEnum/Float64/evaluation_turbo	11.8 ± 1.4 μs	11.5 ± 1.2 μs	1.03
time_to_load	13.1 ± 0.13 s	12.9 ± 0.24 s	1.02

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR. Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

[MilesCranmer]

Can have strange behavior due to the @fastmath: https://discourse.julialang.org/t/fastest-way-to-check-for-inf-or-nan-in-an-array/76954/30?u=milescranmer

So perhaps it should do a normal sum on each segment instead (less likely to overflow if its just 64 numbers).