unify AudioToMelSpectrogramPreprocessor and AudiotoSpectrogram

[krishnacpuvvada]

    this is very close to 

https://github.com/NVIDIA/NeMo/blob/main/nemo/collections/asr/modules/audio_preprocessing.py#L95. it would just be adding one or two args to return intermediate STFT from AudioToMelSpectrogramPreprocessor or sub-classing it.

Originally posted by @krishnacpuvvada in #5356 (comment)