Since your trails are in a separate feature class, as mentioned above, you could construct a single cost attribute that calculates the travel time on roads using driving speeds and on trails using walking speeds. So technically you could get this to work. You just need to set up the evaluators for your cost attribute correctly.
Documentation for cost attributes: https://pro.arcgis.com/en/pro-app/latest/help/analysis/networks/cost-attributes.htm
Documentation for evaluators: https://pro.arcgis.com/en/pro-app/latest/help/analysis/networks/types-of-evaluators-used-by-a-networ...
The limitation with this model is that the solver can't track whether or not the traveler currently has their car. So, you could potentially get a route where the traveler drives to a trailhead, walks on the trail, emerges at a different point, and continues driving, even though they left their car at the other end of the trail. But if you're just trying to find the best way to get to some point deep in the woods from your headquarters or something like that, it shouldn't be an issue.
