Fix Loose Dependency Imports

[abheesht17]

In https://github.com/keras-team/keras-nlp/blob/master/API_DESIGN_GUIDE.md#avoid-new-dependencies,

try:
    import rouge_score
except ImportError:
    pass

class RougeL(keras.metrics.Metric):
    def __init__(self):
        if rouge_score is None:
            raise ImportError(
                'RougeL metrics requires the rouge_score package. '
                '`pip install rouge-score`.')

raises a NameError if rouge_score is not installed. We can change this to

try:
    import rouge_score
except ImportError:
    pass

class RougeL(keras.metrics.Metric):
    def __init__(self):
        if "rouge_score" not in sys.modules:
            raise ImportError(
                'RougeL metrics requires the rouge_score package. '

Also, all the tests fail because we haven't added rouge_score to setup.py. How do we deal with this issue?

[mattdangerw]

Thanks! We should actually just do.

try:
    import rouge_score
except ImportError:
    rouge_score = None

That is simpler than sys.modules.

Re how to do testing, we should add the pip dependency to the extra_require tests block of our setup.py. That will add the install during our ci actions.

[abheesht17]

Ah, yes. The above code snippet is way easier. Will update the PR! Thanks!