Handle conditionals and abs

[ChrisRackauckas]

It's time we finally handle #531, https://github.com/SciML/ModelingToolkit.jl/issues/532, and https://discourse.julialang.org/t/differentiation-method-for-element-wise-abs-function-applied-on-operation-types-in-modelingtoolkit-jl/45867/4 . This PR creates a ModelingToolkit.ifelse with derivative fixes for abs to make it so standard conditional code can work. While tracing cannot correctly make these operations, this would at least allow directly written symbolic code to handle conditions, so we're at least as good (or bad) as TensorFlow.

[ChrisRackauckas]

@chriselrod should we make a IfElse.jl package so that way we use the same one as LoopVectorization.jl? @c42f should we inject this as Base.ifelse?

[ChrisRackauckas]

using ModelingToolkit
function func!(du,u)
    du = abs.(u)
end
@variables du[1:2] u[1:2]
func!(du,u)
sjac= ModelingToolkit.sparsejacobian(vec(du),vec(u));

function func!(du,u)
    for j = 1:2
        du[j] = abs(u[j])
    end
 end
@variables du[1:2] u[1:2]
func!(du,u)
sjac= ModelingToolkit.sparsejacobian(vec(du),vec(u));

works with this PR.

[YingboMa]

LGTM. I agree that it's better to make a new "IfElse.jl" package so that we can overload the same function.