Synthetic Semantic Segmentation Benchmark

[czaloom]

Changes

• Added benchmark utilities to valor-lite.profiling
• Added benchmarks in valor-lite.semantic_segmentation.benchmark
• Added example notebook benchmarking.ipynb

[rsbowman-striveworks]

I think you should separate out the parts that generate synthetic data because those parts are useful outside of benchmarking. For example this could be used by something similar to the test_stability.py unit tests that detection and classification have. I have also used these to play around with the API and see what evaluations look like with different options. Very useful IMO and worth making it easy to do.

Not exactly sure where it should go, maybe in segmentation/synthetic.py or segmentation/test_data.py? Maybe right in segmentation/annotation.py itself? I'm open to other ideas.

[czaloom]

Moved generate_segmentation to annotation.py!