Dataeval

📝 **Miscellaneous**
- `32ba1f29` - Data split tests
- `76f73770` - Updated glossary and other files to use new style of links
- `20efd27e` - Add support for Python 3.12

🌟 **Feature Release**
- `14ef382c` - Update dependencies for conda compatibility

🛠️ **Improvements and Enhancements**
- `97849b01` - Update support for tensorflow >=2.16 with explicit keras v2

📝 **Miscellaneous**
- `85bafa30` - Swap brightness and darkness
- `96a30ad0` - Make optional checks more granular
- `55ca81d6` - Use native int for dict keys for Outliers
- `639e140b` - silence warnings for docs and doctest

🌟 **Feature Release**
- `cdae8a17` - Parallelize existing stats metric functions and introduce dedicated channelstats function

Running statistical analysis functions take significant time against large datasets. Due to the natural parallelism of analyzing individual images, we introduced parallel processing leveraging the `multiprocessing` library to accelerate processing times.

Affected functions:

* `datasetstats`
* `dimensionstats`
* `hashstats`
* `pixelstats`
* `visualstats`

Additionally, `channelstats` was added which performs the functionality of `datasetstats` but only for the functions that support per-channel stat calculation, `pixelstats` and `visualstats`.

📝 **Miscellaneous**
- `552668a0` - Update EDA part 1 tutorial with miscellaneous changes

🛠️ **Improvements and Enhancements**
- `d1cdcda5` - API changes with supporting documentation updates

📝 **Miscellaneous**
- `5ecd4d3a` - expose datasets API
- `6c19bba7` - Make sufficiency args more permissive
- `1bc2d067` - Improving MNIST class
- `d23b3461` - Extract small-scope reusable functions from tools made for prototype Associate[Drift|OOD]withMetadataTutorial notebooks.
- `5bea9512` - remove tf-iogcs-fs

🌟 **Feature Release**
- `71e7ff06` - Integrate labelstats function
- `f40bf0e4` - Redesign stats functions for expansion to per-box, per-channel, and boxratiostats

🛠️ **Improvements and Enhancements**
- `72390edc` - Change input format of balance and diversity to be the same as parity

👾 **Fixes**
- `f598c46a` - Update pytorch to 2.2.0+

📝 **Miscellaneous**
- `b8f0d502` - Create copy on `to_numpy` by default
- `04a71337` - Fix CI docs job to load on build
- `9286f5e8` - Skip or rework MNIST based unit tests
- `704f44e3` - Investigate the use of metadata to help explain observed dataset drifts and OOD examples
- `e25f84f3` - Expose SufficiencyOutput and move class methods to output class
- `742a084c` - Adding algorithm compatability/requirements table
- `7ce85be7` - Misc concept documentation