Add dummy batch for model warmup in svs_warmup function

Merge remote-tracking branch 'public/main'

Update Chinese-Japanese bilingual model path to be espnet/mixdata_svs_visinger2_spkemb_lang_pretrained

Refactor SVS evaluation and move SingMOS functions to svs_eval.py; add pitch interval and chroma entropy

Add demo and run script

fugashi -> kanjiconv and add touhou music

update jp preprocess

Update Japanese text preprocessing and formatting to support natural Japanese text input without spaces between characters

add jp example

Continue: Refactor svs_inference and related functions; Bug fixes and code cleanup

Continue:

Refactor svs_inference and related functions; Bug fixes and code cleanup

Refactor load_song_database to accept config and update data loading logic; add lyric_word_length calculation in create_features

Continue: Add random_select.take_lyric_continuation for singing a syllable with multiple successive pitches

Push demo

unify configs & add readme

Add random_select.take_lyric_continuation for singing a syllable with multiple successive pitches

syntax updates

Fix cache dir bug

Co-authored-by: Yiwen Zhao <Tsukasane@users.noreply.github.com> & Jionghao