Enhancement: Expanded Audio Format Support with Soundfile

[mbsantiago]

This pull request significantly enhances the audio loading capabilities of our soundevent library.

Key Improvements:

• Expanded Format Support:
  • Previously, soundevent only supported WAV files. We've now integrated the soundfile library, enabling support for a wide array of lossless and lossy audio formats (FLAC, MP3, etc.), making the library more versatile and accommodating a broader range of use cases.
  • We excluded a few formats that did not allow reading audio clips without loading the entire file, ensuring efficient processing of large recordings.
• Efficient Metadata Retrieval & Clip Reading: Critical metadata (duration, sample rate, channels, etc.) can be obtained quickly without fully loading the audio file. Additionally, you can now read specific audio clips from a file without loading it entirely. This is crucial when working with large audio files, as it significantly improves performance.
• Standardized Audio Attributes: The audio.attributes module now defines standardized attributes for audio arrays, aligning with the new arrays module and promoting consistency throughout the codebase.
• Consolidated Audio Operations: We've created a dedicated audio.operations module for audio-specific array operations (filtering, resampling, PCEN), resulting in a cleaner code organization.
• Improved PCEN Implementation: The PCEN (Per-Channel Energy Normalization) transform has been reimplemented to adhere more closely to the original paper, eliminating the dependency on librosa.
• Enhanced Audio File Identification: The audio.is_audio_file function now recognizes all supported formats. We've also added a "strict" mode for more thorough file validation using soundfile.
• New Utility Function: A new audio.get_audio_files function allows easy recursive iteration over all audio files within a directory.

Breaking Changes:

• The MediaInfo object has been modified to include format and subtype fields (replacing format and bit_depth). This change was necessary to accommodate the wider range of audio formats. While it's a breaking change, it significantly improves the library's flexibility.

Additional Notes:

• This pull request addresses feedback from @alvarovegahd regarding the importance of FLAC support. (see Enhancement Proposal - Adding FLAC File Support to load_audio Function #1)
• We removed redundant operations and the audio.spectrum module due to the expanded capabilities provided by soundfile and the new arrays.operations module.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.21%. Comparing base (f9a6d1a) to head (fb4943d).

Additional details and impacted files

@@            Coverage Diff             @@
##             main       #8      +/-   ##
==========================================
+ Coverage   86.87%   88.21%   +1.33%     
==========================================
  Files         106      102       -4     
  Lines        3154     2995     -159     
  Branches      492      470      -22     
==========================================
- Hits         2740     2642      -98     
+ Misses        321      283      -38     
+ Partials       93       70      -23     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.