Shuffle result in _fit_resample method

[liyakun]

Is your feature request related to a problem? Please describe

The current method _fit_resample in _random_over_sampler.py simply appends the new samples and returns the result. However, it might be helpful to shuffle the result before returning. For example, in deep learning model training, the data from resampling could result in samples in one batch with same label which could lead to poor performance in training the model.

Describe the solution you'd like

Shuffle the result in _fit_resample before returning it

[hayesall]

That's the motivation in imblearn.keras.BalancedBatchGenerator

Indices are shuffled at the end of the sample_ method:

imbalanced-learn/imblearn/keras/_generator.py

Lines 179 to 180 in 9f8830e

	# shuffle the indices since the sampler are packing them by class
	random_state.shuffle(self.indices_)

[glemaitre]

Closing then.