feat: implement the shortest path algorithm in the system

[merlintang]

• implement the weighted shortest path
• implement the unweighted shortest path
• with limit

MATCH p=shortestPath(
 (start:Person {name: 'Bob'})-[:KNOWS*]-(end:Person {name: 'Fleur'}))
RETURN length(p) AS length

[merlintang]

this part would be one of our killing functions, can we optimize this to compete with neo4j or others?

from my point of view, this is important for our applications and good for advisertisement.

@yangke110 @zhiyuan-yang

[merlintang]

let us reuse the approaches here.

https://github.com/iwiwi/pruned-landmark-labeling

https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.739.4799&rep=rep1&type=pdf

[merlintang]

build the index, and then do the breadth-first to find the shortest path.

before we implement this, we need to consider how can we do it via the naive approach.

[merlintang]

https://github.com/Amtrix/parallel-pruned-landmark-labeling the parallel version.

[merlintang]

https://github.com/y3eadgbe/PrunedLabeling

[merlintang]

let us think about this again under the system context.

1. how can we do it in the iterator model or vector model
2. how can improve this in the k-v storage format
3. how can hanlde the index update or etc.