Breadth-first search

[Robbybp]

When exploring a new model, I like doing a limited-depth BFS starting from the variables, constraints, or objective that I care about. This PR adds methods for a limited-depth BFS. Here's an example:

import PGLib
import PowerModels
import MathProgIncidence as MPIN
import JuMP
data = PGLib.pglib("14")
pm = PowerModels.instantiate_model(data, PowerModels.ACPPowerModel, PowerModels.build_opf)
tree = MPIN.limited_bfs(JuMP.objective_function(pm.model); depth = 2, include_inequality = true)
println(tree)

2326.9494 0_pg[2] + 792.0951 0_pg[1]
├ 0_pg[2]
│ ├ -0_pg[2] + 0_p[(5, 2, 5)] + 0_p[(3, 2, 3)] + 0_p[(4, 2, 4)] + 0_p[(1, 2, 1)] = -0.217
│ ├ 0_pg[2] ≥ 0
│ └ 0_pg[2] ≤ 0.59
└ 0_pg[1]
│ ├ -0_pg[1] + 0_p[(1, 1, 2)] + 0_p[(2, 1, 5)] = 0
│ ├ 0_pg[1] ≥ 0
│ └ 0_pg[1] ≤ 3.4

TODO

• Tests
• Maybe slightly better printing