Add KKT solver: cliquetrees.

[samuelsonric]

I implemented a fast LDLt factorization routine in Julia.

https://github.com/AlgebraicJulia/CliqueTrees.jl

This PR adds it to Clarabel.jl. It performs similarly to the MA57 solver. For example:

using Clarabel, CliqueTrees, FileIO, JuMP, LinearAlgebra, HSL

data = load("./SDPLIB/mcp250-2.jld2");
F = data["F"];
c = data["c"];
m = data["m"];
n = data["n"];
obj_true = data["optVal"];

model = JuMP.Model(Clarabel.Optimizer);
set_attribute(model, "direct_solve_method", :cliquetrees)
set_attribute(model, "verbose", false)
@variable(model, x[1:m]);
@objective(model, Min, c' * x);
@constraint(model, con1,  Symmetric(-Matrix(F[1]) + sum(Matrix(F[k + 1]) .* x[k] for k in 1:m)) in JuMP.PSDCone());
@time JuMP.optimize!(model);
objective_value(model)

Here are the times for the different Julia solvers.

solve	time (s)
cliquetrees	10.4
ma57	12.0
qdldl	58.1
cholmod	63.0