Refactor node ordering logic

[hdevalence]

The logic here could be simplified using a tuple comparison. Also it's not required to duplicate logic between the PartialOrd and Ord impls, Ord is a marker trait. https://github.com/ZcashFoundation/zebra/blob/master/zebra-p2p/src/util/node_table.rs#L54-L92

[hdevalence]

note also: because the NodeTable maintains BTreeSets for by_score, by_time, but the custom-ordering-tuple-newtypes don't hold borrows, all of the node data is held in triplicate

[hdevalence]

aha, we may not be able just to simplify the comparison functions, because the code that maintains the data in triplicate stops working if we relax requirements on the sort-by-x functions (e.g., so that two nodes with equal time are equal when sorted by time).