Ricciardi

Minor update:
- slightly more optimized `ricciardi` function
- slightly better documentation.

**Full Changelog**: https://github.com/hchau630/ricciardi/compare/v0.1.4...v0.1.5

- Add support for integer tensor inputs
- More elegant NaN gradient handling for large negative inputs resulting in improved performance

**Full Changelog**: https://github.com/hchau630/ricciardi/compare/v0.1.3...v0.1.4

- Minimum supported Python version set to 3.9
- Added support for torch.bfloat16
- Added type hints

**Full Changelog**: https://github.com/hchau630/ricciardi/compare/v0.1.2...v0.1.3

- Add precision level option to `ricciardi` and appropriately handle half or double precision inputs.

- Removed a (somewhat) unnecessary `.double()` call for improved performance with a very slight accuracy tradeoff.