Marking as on hold because I need to add more unittests to improve the coverage, and also add more than the simple cases.

Pull Request Test Coverage Report for Build 355266041

• 47 of 47 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.1%) to 94.229%

Totals
Change from base Build 354704726:	0.1%
Covered Lines:	2596
Relevant Lines:	2755

💛 - Coveralls

Similar to #147, I think it would be better to have insert_node_on_in_edges(node, node_ref) and insert_node_on_out_edges rather than unified insert_node_between(node, node_ref, direction). What do you think?

Yeah that's a good idea. I'll update it shortly to split the functions

@mtreinish Thank you. I have one more comment on docstrings.

I think we can describe the spec more clearly without parent/child notation:

/// Insert a node between a reference node and all its predecessor nodes
///
/// This essentially iterates over all edges into the reference node specified
/// in the ``ref_node`` parameter removes those edges and then adds 2 edges,
/// one from the predecessor of ``ref_node`` to ``node`` and the other from
/// ``node`` to ``ref_node``. The edge payloads for the newly created edges
/// are copied by reference from the original edge that gets removed.
///
/// :param int node: The node index to insert between
/// :param int ref_node: The reference node index to insert ``node`` between
pub fn insert_node_on_in_edges(..., node, ref_node)

/// Insert a node between a reference node and all its successor nodes
///
/// This essentially iterates over all edges out of the reference node specified
/// in the ``ref_node`` parameter removes those edges and then adds 2 edges,
/// one from ``ref_node`` to ``node`` and the other from ``node`` to the
/// successor of ``ref_node``. The edge payloads for the newly created edges
/// are copied by reference from the original edge that gets removed.
pub fn insert_node_on_out_edges(..., node, ref_node)

(I like the term reference node (ref_node) while I'm afraid someone might confuse with "reference of node object" (node_ref). Anyway, I think we could have better name than node_between. Any good idea?)

BTW, I personally prefer start -> end notation to parent -> child or tail -> head notation.
Honestly, I'm always confused with which is the "head" of parent and child...
(I know the importance of consistency, so if we have already used parent/child notations in many places, I think we should follow the convention.)

@itoko thank you for the suggestions. I was actually hoping you'd leave a comment like that. I thought the names were awkward in this function, but couldn't come up with anything better so just left it as is. I'll update it now per your suggestions.