Recovery simulations

[annehaley]

This PR adds a new Simulation type, "Recovery Scenario". For its input arguments, the user can select the following:

• an existing SimulationResult object from a node failure simulation
• a recovery mode from a list of strings: random, betweenness, degree, information, eigenvector, load, closeness, second order

The implementation of this simulation utilizes a function written by Jack Watson at NEU, shared with us for this simulation. The options available in that function determine the list of available recovery modes.

The animation of the results of this simulation only required a slight modification to the NodeFailureAnimation component, now renamed NodeAnimation. The component now accepts a list of node_recoveries alongside the list of node_failures. If node_recoveries is specified, the animation will subtract the current slice of recovered nodes from the list of failed nodes.

[jjnesbitt]

@annehaley I pushed 476273f, I hope you don't mind. I noticed that in construct_edge_list, node.adjacent_nodes was being called once per node, causing a SQL query to be made for every node in the graph. Since the amount of nodes we currently have is only in the hundreds, and we only run this app locally, those n sql queries don't have a humongous impact. However, if either of those two situations were to change, that function would become very slow indeed. I updated that function to instead make 2 SQL queries, which seems to perform much better, and should be able to handle much larger graphs.

I believe I tested it enough to show it has feature parity with the existing implementation, but please give it a try to make sure I'm not missing anything, and/or otherwise lmk what you think.

[annehaley]

@AlmightyYakob I tried out 476273f and it works well. Thank you for optimizing that.