Fix memory issue in MOI's interface

[nrontsis]

_scalecoef was called (by scalecoef) with a wrong parameter d for PositiveSemidefiniteConeTriangle sets. Instead of d=n where n is the number of rows/columns of the PSD matrix, it was called with n*(n+1)/2. Furthermore _scalecoef was memory inefficient.

As a result the MOI interface was having memory issues when n > 1000 (see example below).

This PR avoids this issue by a different implementation of _scalecoef which does not need d.

Example

Download SDPLIB's equalG51.jld2.zip and compare the execution time & memory requirements of the MOI wrapper for the following code before and after the PR

using LinearAlgebra, SparseArrays
using COSMO, JuMP
using JLD2

function solve_sdpa_jump_dual(c, F; solver=COSMO.Optimizer, kwargs...)
    model = Model(with_optimizer(solver; kwargs...))
    m = length(c); n = size(F[1], 1)
    @variable(model, Y[1:n, 1:n], PSD)
    for i = 1:m
        @constraint(model, dot(Y, F[i + 1]) == c[i])
    end
    @objective(model, Max, dot(F[1], Y))

    JuMP.optimize!(model)
    @show JuMP.termination_status(model)
    return value.(Y), JuMP.objective_value(model)
end

@load "equalG51.jld2" c F blocks
solve_sdpa_jump_dual(c, F[:, 1])

[codecov-io]

Codecov Report

Merging #87 into master will decrease coverage by 0.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master      #87      +/-   ##
==========================================
- Coverage   90.09%   90.08%   -0.02%     
==========================================
  Files          22       22              
  Lines        1707     1704       -3     
==========================================
- Hits         1538     1535       -3     
  Misses        169      169

Impacted Files	Coverage Δ
src/MOIWrapper.jl	92.34% <100%> (-0.07%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cc32a30...b26da48. Read the comment docs.

[migarstka]

Thanks Nikitas! That indeed speeds up the problem setup by a lot.