feat: extra simd kernel branches

[MilesCranmer]

This creates SIMD kernels for op(op2(x, y), z) and op(x, op2(y, z)) for the cases where x, y, and z are all leaf nodes.

[github-actions]

Benchmark Results

	master	6df75c0...	master/6df75c02d0efe1...
eval/ComplexF32/evaluation	7.24 ± 0.54 ms	7.32 ± 0.47 ms	0.99
eval/ComplexF64/evaluation	10.7 ± 0.8 ms	10.7 ± 0.81 ms	0.997
eval/Float32/derivative	11.7 ± 1 ms	11.7 ± 0.9 ms	1
eval/Float32/derivative_turbo	11.7 ± 1.1 ms	11.7 ± 0.97 ms	1
eval/Float32/evaluation	2.7 ± 0.24 ms	2.74 ± 0.26 ms	0.988
eval/Float32/evaluation_bumper	0.613 ± 0.018 ms	0.607 ± 0.016 ms	1.01
eval/Float32/evaluation_turbo	0.544 ± 0.037 ms	0.553 ± 0.037 ms	0.984
eval/Float32/evaluation_turbo_bumper	0.615 ± 0.019 ms	0.608 ± 0.018 ms	1.01
eval/Float64/derivative	15.4 ± 0.75 ms	14.9 ± 0.66 ms	1.03
eval/Float64/derivative_turbo	15.8 ± 0.88 ms	14.9 ± 0.58 ms	1.06
eval/Float64/evaluation	3.23 ± 0.31 ms	3.17 ± 0.29 ms	1.02
eval/Float64/evaluation_bumper	1.27 ± 0.046 ms	1.26 ± 0.042 ms	1.01
eval/Float64/evaluation_turbo	1.05 ± 0.077 ms	1.03 ± 0.069 ms	1.02
eval/Float64/evaluation_turbo_bumper	1.27 ± 0.049 ms	1.26 ± 0.046 ms	1.01
utils/combine_operators/break_sharing	0.0384 ± 0.00063 ms	0.0389 ± 0.00091 ms	0.988
utils/convert/break_sharing	27.4 ± 3.2 μs	27.5 ± 2.7 μs	0.997
utils/convert/preserve_sharing	0.0972 ± 0.0047 ms	0.0989 ± 0.0051 ms	0.982
utils/copy/break_sharing	27.5 ± 3 μs	27.7 ± 2.6 μs	0.995
utils/copy/preserve_sharing	0.0977 ± 0.0046 ms	0.099 ± 0.0048 ms	0.987
utils/count_constant_nodes/break_sharing	8.8 ± 0.28 μs	8.7 ± 0.31 μs	1.01
utils/count_constant_nodes/preserve_sharing	0.0851 ± 0.0039 ms	0.0854 ± 0.0044 ms	0.997
utils/count_depth/break_sharing	9.96 ± 0.56 μs	9.1 ± 0.26 μs	1.09
utils/count_nodes/break_sharing	8.44 ± 0.24 μs	8.77 ± 0.21 μs	0.962
utils/count_nodes/preserve_sharing	0.0842 ± 0.0037 ms	0.0859 ± 0.0045 ms	0.981
utils/get_set_constants!/break_sharing	0.0336 ± 0.0027 ms	0.0335 ± 0.0029 ms	1
utils/get_set_constants!/preserve_sharing	0.175 ± 0.006 ms	0.177 ± 0.0064 ms	0.987
utils/get_set_constants_parametric	0.0443 ± 0.0026 ms	0.045 ± 0.0028 ms	0.984
utils/has_constants/break_sharing	4.08 ± 0.18 μs	3.89 ± 0.13 μs	1.05
utils/has_operators/break_sharing	2.02 ± 0.076 μs	1.86 ± 0.052 μs	1.09
utils/hash/break_sharing	22.6 ± 0.67 μs	23.1 ± 0.62 μs	0.977
utils/hash/preserve_sharing	0.0967 ± 0.0038 ms	0.0981 ± 0.0046 ms	0.985
utils/index_constant_nodes/break_sharing	24.9 ± 1.1 μs	24.7 ± 1.3 μs	1.01
utils/index_constant_nodes/preserve_sharing	0.0971 ± 0.0043 ms	0.101 ± 0.0055 ms	0.958
utils/is_constant/break_sharing	4.35 ± 0.16 μs	4.13 ± 0.11 μs	1.05
utils/simplify_tree/break_sharing	0.17 ± 0.0044 ms	0.17 ± 0.0058 ms	0.997
utils/simplify_tree/preserve_sharing	0.225 ± 0.007 ms	0.222 ± 0.008 ms	1.01
utils/string_tree/break_sharing	0.467 ± 0.025 ms	0.464 ± 0.018 ms	1.01
utils/string_tree/preserve_sharing	0.583 ± 0.028 ms	0.57 ± 0.025 ms	1.02
time_to_load	0.241 ± 0.008 s	0.231 ± 0.0027 s	1.05

[coveralls]

Pull Request Test Coverage Report for Build 12407763507

Details

• 395 of 409 (96.58%) changed or added relevant lines in 3 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.1%) to 95.671%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/Evaluate.jl	198	212	93.4%

Totals
Change from base Build 12322890369:	0.1%
Covered Lines:	2939
Relevant Lines:	3072

---

💛 - Coveralls