fix: Resolve detect Command Failure with Specific Audio Files (GH-31)

[mbsantiago]

This PR addresses issue #31, where the detect command failed to process a set of 5 audio recordings provided by @Jeff37.

The root cause was traced to the batdetect2.utils.audio_utils.pad_audio function, which incorrectly handled cases with short audio inputs. This function prepares audio data for the batdetect2 model by padding it to ensure the resulting spectrogram has dimensions compatible with the model's downsampling and upsampling layers.

To resolve this:

• Refactored pad_audio for improved clarity and correctness, introducing helper functions for spectrogram width calculations and audio sample conversions.
• Added unit tests for pad_audio and the new helper functions to ensure proper functionality.
• Included an integration test that runs the detect command on the problematic audio files provided by @Jeff37, verifying the fix.

All tests now pass, confirming that the issue has been resolved.

Key changes in this PR:

• Refactored pad_audio function in batdetect2.utils.audio_utils.
• Added unit tests for pad_audio and related helper functions.
• Created an integration test using the user-provided audio files to reproduce and verify the fix for issue Error processing file!: negative dimensions are not allowed #31.

Acknowledgments:

Many thanks to @Jeff37 for reporting this issue and providing the valuable test files.