Release v1.0.4

• Fixed: preserve order of segments in index
  after augmentation with auglib.Augment.augment()
• Fixed: ensure transforms work with signals
  stored in a read-only numpy.array
• Fixed: return correct float types with numpy>=2.0

Release v1.0.3

• Fixed: broken links
  on the landing page
  of the documentation

Release v1.0.2

• Fixed: auglib.transform.PinkNoise
  when applied to signals with odd sample length

Release v1.0.1

• Fixed: missing audio examples in documentation
• Fixed: add missing entries to the changelog of 1.0.0

Release v1.0.0

• Added: public release
• Added: support for MacOS
• Added: support for Windows
• Added: audio examples and figures
  to the docstrings
  of all transforms
• Added: depend on scipy
• Added: sampling_rate as argument
  to __call__() of transforms
  that require a sampling rate
• Added: support for Python 3.9, 3.10 and 3.11
• Changed: auglib.Augment.augment()
  now stores augmented segments
  as single files in cache
• Changed: auglib.Augment.augment()
  now uses allow_nat=True
  when calculating the hash of an index,
  which is used to store/find augmented files
  in cache
• Changed: depend on audeer>=2.0.0
• Changed: convert code base to pure Python
• Changed: switch to MIT License
• Fixed: support overlapping segments
• Fixed: ensure the duration argument
  of auglib.transform.AppendValue
  and auglib.transform.PrependValue
  is first observed
  when calling the transform
• Removed: sounddevice dependency