BatchFeature should cast to np.float32 by default

[patrickvonplaten]

Currently the default dtype for Speech Feature Extractors is numpy.float64 which leads to two problems:

1. It makes the data processing extremely expensive for the RAM. Many sound formats are stored in int16 (such as .wav) and are then transformed to float64 which unnecessarly increases RAM by a factor of 4. We should at least stick to float32
2. Currently we have added some hacks to the Wav2Vec2 and Speech2TextTransformer feature extractors to prevent Double vs. Float dtype mismatches:

   transformers/src/transformers/models/wav2vec2/feature_extraction_wav2vec2.py

   Line 87 in f6e2544

   input_values = [x.astype(np.float32) for x in input_values]

The main problem is that np.asarray([....]) by default creates a np.float64 array and that we just pass that format along.
=> We should either always cast to float32 in BatchFeature (see here:

transformers/src/transformers/feature_extraction_utils.py

Line 151 in f6e2544

as_tensor = np.asarray

) or add a flag dtype to BatchFeature.

@patrickvonplaten