Implement a tf.version check

[LukeWood]

https://github.com/keras-team/keras-tuner/blob/master/keras_tuner/utils.py#L57

We need 2.8.0>

[bhack]

The problem is also that depending only on tf-nightly doesn't enforce anything if we don't define > date
e.g: tf-nightly 2.9.0.dev20220322.

Currently with only tf-nightly will mean that any nightly dependency satisfy keras-cv and so also a pre 2.8.0 or a 2 years old tf-nightly

https://github.com/keras-team/keras-cv/blob/master/setup.py#L35

[LukeWood]

This is done.