Tf-explain

What's Changed
* Ensure that input tensors are watched in Grad-CAM by craymichael in https://github.com/sicara/tf-explain/pull/167
* Update usage.rst by andife in https://github.com/sicara/tf-explain/pull/170
* Add TensorFlow versions 2.5 and 2.6 in CI by RaphaelMeudec in https://github.com/sicara/tf-explain/pull/175

New Contributors
* craymichael made their first contribution in https://github.com/sicara/tf-explain/pull/167
* andife made their first contribution in https://github.com/sicara/tf-explain/pull/170

**Full Changelog**: https://github.com/sicara/tf-explain/compare/v0.3.0...v0.3.1

Release includes:

- Migration from TravisCI to Github Actions
- Compatibility with TensorFlow 2.3 and 2.4
- Fix in Vanilla Gradients computation (use raw scores instead of softmax outputs)
- Let user choose between opencv versions

In this release:

- [x] Checked against TensorFlow 2.1.0
- [x] `image_weight` parameter added to GradCAM to better handle visualizations
- [x] Grids of attribution maps are now normalized image by image instead of the whole grid
- [x] bugfix : reshape of non-square visualization by jpsimen

In this release:

- [x] Gradients*Inputs algorithm by ywolff
- [x] Add layer detection for Grad CAM (last convolutional output) to ease user experience with this method
- [x] Better handling of image output ([0, 1], [-1, 1] & [0 255])
- [x] Notebooks

In this release:

- [x] TensorFlow 2.0.0 compatibility
- [x] Add Vanilla Gradients implementation
- [x] Add Integrated Gradients implementation
- [x] Take Viridis as default colormap

Thanks to ywolff for the contributions in this release.

In this release:

- [x] Add SmoothGrad implementation
- [x] Improve computation time with `tf.function` decorator
- [x] Tensorboard integration
- [x] Docstring & Testing improvements