Add dijkstra all_pairs functions

[mtreinish]

This commit adds 2 new api functions all_pairs_dijkstra_shortest_paths()
and all_pairs_dijkstra_path_lengths(). These two functions are similar
to the existing dijkstra functions but they differ in that it will
return dictionaries over all nodes in the graph instead of just for a
single user provided node. Additionally these functions are multihreaded
and the level of parallelism can be controlled by the appropriate
environment variables.

Fixes #271

[coveralls]

Pull Request Test Coverage Report for Build 865105436

• 316 of 352 (89.77%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-1.3%) to 95.502%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/iterators.rs	316	352	89.77%

Totals
Change from base Build 864718571:	-1.3%
Covered Lines:	5669
Relevant Lines:	5936

---

💛 - Coveralls

[mtreinish]

I ran some benchmarks last night on this:

import time

import retworkx as rx
import networkx as nx

rx_graph = rx.directed_gnp_random_graph(5000, .9)
nx_graph = nx.DiGraph(list(rx_graph.edge_list()))

start = time.time()
rx.digraph_all_pairs_dijkstra_shortest_paths(rx_graph, lambda _: 1)
stop = time.time()
print("retworkx: %s" % str(stop - start))

nx_start = time.time()
# Need to cast as a dict because networkx returns an iterator
dict(nx.all_pairs_dijkstra_path(nx_graph))
nx_stop = time.time()
print("networkx: %s" % str(nx_stop - nx_start))

Library	Time (sec)
retworkx	93.04283165931702
networkx	14554.81346654892

[nahumsa]

Great implementation overall. Using AllPairsPathLenghtMapping and AllPairsPathMapping makes it a clean implementation. Just had a few questions regarding the docs.