Addition of copy_model and copy_gdp_data to copy GDP models.

[dnguyen227]

This PR extends JuMP.copy_extension_data in order to allow GDPModels to be an input to JuMP.copy_model().
JuMP.copy_extension_data does not receive a reference map as one of it's argument so it's impossible to recover disjunct constraints. Therefore another function was made to be called (DP.copy_gdp_data()) in order to populate the new model's GDPData. The workflow goes sequentiallty as in the example below.

using HiGHS, JuMP
import DisjunctiveProgramming as DP  
m = DP.GDPModel(HiGHS.Optimizer)
@variable(m, 0 ≤ x[1:2] ≤ 20)
@variable(m, Y[1:2], DP.Logical)
@constraint(m, [i = 1:2], [2,5][i] ≤ x[i] ≤ [6,9][i], DP.Disjunct(Y[1]))
@constraint(m, [i = 1:2], [8,10][i] ≤ x[i] ≤ [11,15][i], DP.Disjunct(Y[2]))
DP.@disjunction(m, Y)
@objective(m, Max, sum(x))

m_copy, ref_map = JuMP.copy_model(m)
lv_map = DP.copy_gdp_data(m, m_copy, ref_map)
set_optimizer(m, HiGHS.Optimizer)
set_optimizer_attribute(m, "output_flag", false) 

set_optimizer(m_copy, HiGHS.Optimizer)
set_optimizer_attribute(m_copy, "output_flag", false)

optimize!(m, gdp_method = DP.BigM())
optimize!(m_copy, gdp_method = DP.Hull())

In addition there is also copy_model_and_gdp_data to do both
An alternative would be to extend JuMP.copy_model itself but at a glance I believe there might be consequences when we want to have DP coexist with other packages and have the ability to copy.

[dnguyen227]

@pulsipher this is ready for review.

[pulsipher]

Looking good, just some minor points to address

[pulsipher]

Looks good, just one more thing. Please add some comments make the code more readable. For example, a comment for each for loop in copy_gdp_data and some comments for the dispatch functions. Also, the new public functions should each have a docstring.

[pulsipher]

Looks good to me.