Computes the spectrogram of an audio signal.

Args:
audio_tensor: Input audio tensor of shape [samples] or [channels, samples]
config: Spectrogram configuration options NxAudio.Transforms.SpectrogramConfig

Returns:
If input is [samples]: Returns tensor of shape [time, frequency]
If input is [channels, samples]: Returns tensor of shape [channels, time, frequency]

Note: Expected input format is {channels, time} for multi-channel audio