Build cross-vpu nexus dataset.

To support a variety of use cases, the high-res mainstems dataset will need to adhere to a clean domain decomposition that is compatible with the input data packaging. Four-digit HUCs are used as the domains in the case of practically all NHD products and are the obvious domains to use.

Each nexus will be identified by the reference mainstem that it lies along and the four digit huc it drains. e.g. "16996670315" is the nexus that recieves flow from HUC 0315 along mainstem 1699667.

HUC04 units are greater than 10k sqkm and have rivers draining them that are several hundred km long. In many cases, the fine-grained detail of confluence network topology implies an upstream downstream relationship where the "flow through" distance is on the order of meters or a km. In cases like this, it is important that the fine-grained topology is maintained but referenced to a more accurate flow network topology that abstracts the local detail into a regional network. To accommodate this, only one cross-vpu nexus will exist for a confluence that may imply additional cross-vpu relationships. For example, where the Yellowstone river flows into the Missouri river, the Yellowstone enters the Missouri just upstream of where the Missouri enters the next downstream HUC04 (shown in screenshot below). This is all one confluence but will appear as a more nuanced set of network relationships in tables for the various hydrographies that represent it.

Yellowstone flows in from the south east. Missouri flows through from the west to the east. Green dots are where the NHDHR features break between VPUs. Red circle is an additional location where NHDPlusHR crosses between VPUs in a slightly different discritization.