fzyukio/koe: April 2021 Release

New stuff Spectrograms are now paged (stored separately in folders not exceeding 1000 spectrograms each) so the load time is vastly improved Features can now be aggregated by min, max, var, begin and end values Added parameters to database model which can be set at the creation of the database, or changed in database management Change feature extraction to use database's specific NFFT and overlap Spectrogram preferences (colour map, zoom, etc...) are now saved - no need to change them each time you open a new file Spectrogram now has y-axis Added script to export koe syllables to Excel, including spectrogram Apply bandpass filter in feature extraction Make Koe compatible with Python 3.9, Mac Big Sur, and Apple Silicon (must use conda-forge) Koe's docker images are now truly handsfree, no more messing with project settings on-site Fixed stuff Fixed various problems with spectrograms not being generated when importing CSV Fix inconsistency in file's sampling rate across: original WAV, converted MP3 and Browser's default sampling rate Use 2 decimal points for segment start and end times Fix bug: end_ms > duration Fix various bugs when upload raw recording Fix fundamental frequency not returning correct value when the database has bandpass filte