Convert from daggy to petgraph stabledigraph

This commit migrates the internal graph representation of a PyDAG object
from daggy to use stabledigraph in petgraph directly. This removes an
abstraction layer from the stack and also makes it easier to use the dag
in python since the indexes are stable now. The implementation of the
petgraph trait implementations is borrowed heavily from
mitchmindtree/daggy#21 which added a stable
graph dag implementation to daggy. If/when daggy pushes a new release
that includes this implementation and keeps up to date with the latest
version of petgraph it will probably be worth switching back to that
newer version of daggy.

The current issue is most of the petgraph algorithms do not work with
StableGraph types and instead rely solely on Graph types. As a
workaround for this we will need to add local implementations of these
algorithms in the repo. Then after we fix these issues upstream and
include them in a release we can switch back to petgraph's
implementation.

Also, right now the scaling performance of this is limited by the cycle
detection. At sizes above 10,000 nodes it will take more than a second
to add a child, parent, or edge. To avoid this we can either remove the
runtime cycle detection on every add operation. Or alternatively looking
into adapting the rolling cycle detection we were previously getting for
free from daggy.

Fixes #6

Author: mtreinish

Cargo.toml

@@ -16,9 +16,8 @@ name = "retworkx"
 crate-type = ["cdylib"]
 
 [dependencies]
-petgraph = "0.4"
-daggy = "0.6"
+petgraph = "0.5"
 
 [dependencies.pyo3]
-version = "0.8.3"
+version = "0.8.5"
 features = ["extension-module"]