We've previously demonstrated the flood outage simulation workflow on the MBTA network, but now we should ensure that it works on a much denser network like the Boston Road Network. My initial attempt at this showed that there are several areas of improvement:

1. The load_roads task needs to be updated to create a Layer and LayerFrame for the new data
2. The get_summary function needs to be updated to be compatible with the data from the road network (allow list values for feature properties)
3. The GCC endpoint should be switched to a POST request instead of a GET request. On the denser network, the excluded_nodes list becomes too long to reasonably fit into query parameters, so it should be sent in the request body instead.
4. Running the flood outage simulation took a long time (8 hours for 11387 nodes evaluated for 99 flood frames). There are probably some performance improvements that can be made in the task, computing the pixel region for each station before iterating through the flood frames.
5. Once the simulation is completed, the output animation is unusable because the GCC request made for each tick of the animation takes an unacceptable length of time (7 minutes for a single GCC request). We could circumvent this by precomputing the GCC for each tick of the animation and include that in the task output. But we should also spend the effort to optimize the GCC endpoint, because we will still need it for when the user manually makes node outage states.
6. The Networks panel will need performance improvements, too. The request to get all nodes and the request to get all edges take a while, and this could probabbly be fixed with pagination (and perhaps an infinte scroll component).