Improving routing-only datastream outputs

Creating this issue to document problems with the routing-only datastream outputs and potential solutions. Currently, the routing-only datastream is only running short-range forecasts for VPU03W.

- Very strange flowpaths
This is a screenshot of the latest T+18 routing-only short-range forecast for Alabama

And this is a screenshot of the latest T+18 CFE short-range forecast for Alabama.

Both are quite incorrect in their own ways, but at least the CFE forecast shows the major waterways (Tombigbee, Black Warrior, Alabama, Mobile, Coosa, Pascagoula) very clearly. I'm not sure why the routing-only datastream has such weirdly disjointed flowpaths. Maybe my method to convert NHD flowpaths to the NextGen hydrofabric is flawed? The method is detailed here https://github.com/CIROH-UA/t-route/issues/12

It is also possible everything will stabilize over time, and an 18-hour forecast is simply not long enough.

- Underprediction of flow
This just seems like us not having the correct initial conditions and starting everything at 0.

Here's a segment of the latest T+18 routing-only short-range forecast for the Black Warrior River. According to USGS gage observations, the flow value should be around 330 m^3/s, but here it is only about 2 m^3/s.
As a bonus little treat, this screenshot is another great example of strange flowpaths.

Start by reviewing the NHD-to-NextGen hydrofabric conversion method linked to t-route issue #12, then compare the routing-only T+18 screenshots with the CFE forecast and USGS gage observations described here. Done means identifying and correcting the causes of disjointed flowpaths and severe flow underprediction in the routing-only datastream.

Written by the indexing model from the issue text.