[FEA] Support sample_incoming_edges and sample_outgoing_edges to uniform neighborhood sampling

[VibhuJawa]

Describe the solution you'd like and any additional context

We should support sample_incoming_edges and sample_outgoing_edges for the uniform neighborhood sampling algorithm.

Currently we only support sampling for outgoing edges .

For DGL the default behavior is to sample incoming edges instead of outgoing edges.

From the docs see below:

For each node, a number of inbound (default) (or outbound when edge_dir == 'out') edges will be randomly chosen. The graph returned will then contain all the nodes in the original graph, but only the sampled edges.

We also discussed potentially reversing the edges (See #2458 ) to get the incoming edges but adding the direction to the API might be easier and cleaner.

CC: @ChuckHastings , @seunghwak

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[ChuckHastings]

This can be accomplished today by creating the graph in the opposite direction. Do we need to address this at the C++ layer, or is it sufficient to add logic in some python layer to create the graph with store_transposed=True instead of store_transposed=False?

@alexbarghi-nv ?