Holistictraceanalysis

Added
- (Experimental) Added CUPTI Counter analyzer feature to parse kernel and operator level counter statistics.

Changed
- Improved loading time by parallelizing reads in teh `create_rank_to_trace_dict` function.

Fixed
- Fix unit of time in `get_gpu_kernel_breakdown` API.
- Optimized multiprocessing strategy to handle OOMs when process pool is too large.

Changed
- Split requirements.txt into two files: `requirements.txt` and `requirements-dev.txt`. The former
does not contain `kaleido`, `coverage` as they are required for development purposes only.

Removed
- Coverage tests for graph visualization
- Dependency on Matplotlib for Venn diagram generation

Fixed
- LICENSE type in setup.py
- Queue length summary handles missing data and negative values
- Use `.get` for key lookup in dictionary
- Typos in README.md

- Initial release

- Test release

- Test release