fetch_adjacencies(): Fast-path for simple cases

[stuarteberg]

In simple cases involving explicit lists of relatively few bodies, fetch_adjacencies() has high latency due to extra calls to neuprint. Common simple cases could be special-cased to speed them up.

The fetch_adjacencies() function is designed to perform well (avoid timeouts) even when the source or target NeuronCriteria involves thousands of bodies. To support that, it sends some initial queries to assess the size of the source/target body lists and then pre-fetch the complete body list for one of them. But in cases where the body list is explicitly provided by the user, we could implement a special fast-path that avoids those extra calls. That would improve interactive latency for such calls.