Restore specialized Array SIMD method for calculate_residuals!

The hand-written @simd ivdep loop for Array + Serial() bypasses
FastBroadcast overhead for the common case of plain Arrays with
serial execution.

Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: ChrisRackauckas

lib/DiffEqBase/src/calculate_residuals.jl

@@ -114,6 +114,16 @@ with multiple threads.
     return nothing
 end
 
+@inline function calculate_residuals!(
+        out::Array, ũ::Array, u₀::Array, u₁::Array, α::Number,
+        ρ::Number, internalnorm::F, t, ::Serial
+    ) where {F}
+    @inbounds @simd ivdep for i in eachindex(out, ũ, u₀, u₁)
+        out[i] = calculate_residuals(ũ[i], u₀[i], u₁[i], α, ρ, internalnorm, t)
+    end
+    return nothing
+end
+
 """
     calculate_residuals!(out, u₀, u₁, α, ρ, thread=Serial())