[REVIEW] Profile graph creation runtime and memory footprint

[VibhuJawa]

This PR closes part of https://github.com/rapidsai/graph_dl/issues/187 and #3439
Command to run:

python cugraph_graph_creation.py

See below benchmarks on SG (Works on MNMG too) :

scale	num_input_edges	directed	renumber	input_memory_per_worker	peak_allocation_workers	input_to_peak_ratio	output_to_peak_ratio
22	67108864	TRUE	TRUE	1.0GB	5.9GB	5.885	7.69
22	67108864	TRUE	FALSE	1.0GB	5.9GB	5.885	7.69
22	67108864	FALSE	TRUE	1.0GB	12.1GB	12.08	6.13
22	67108864	FALSE	FALSE	1.0GB	12.1GB	12.08	6.13
23	134217728	TRUE	TRUE	2.0GB	11.8GB	5.914	7.70
23	134217728	TRUE	FALSE	2.0GB	11.8GB	5.914	7.70
23	134217728	FALSE	TRUE	2.0GB	24.3GB	12.16	6.14
23	134217728	FALSE	FALSE	2.0GB	24.3GB	12.16	6.14
24	268435456	TRUE	TRUE	4.0GB	23.8GB	5.93	7.72
24	268435456	TRUE	FALSE	4.0GB	23.8GB	5.93	7.72

[rlratzel]

Nice!

[rlratzel]

/merge