release: prepare FunASR 1.4.11

[LauraGPT]

Summary

• publish the punctuation/timestamp alignment fix from fix(auto-model): align SentencePiece word markers #3587 as FunASR 1.4.11
• update all four top-level language READMEs with the exact abnormal/control sample evidence
• keep the package version contract aligned with the release

Verification

• python -m pytest -q tests/test_punc_model_none.py tests/test_cli.py tests/test_generate_subtitle.py tests/test_release_version_contract.py tests/test_release_package_data.py -> 48 passed
• two exact-commit builds with SOURCE_DATE_EPOCH and normalized sdist are byte-identical:
  • wheel SHA-256 077da7b5e995d5ae13f58396e9fdfe9973e6567a76f77087c6f64fa08867079c
  • sdist SHA-256 e464e64965b28ddcabd4090701c2be18a107a3b017ad3826bde616357225394d
• twine check passes for both artifacts
• clean no-deps wheel install reports version 1.4.11 and contains the SentencePiece marker fix
• reporter's 633.651 s sample: 275 timestamped sentences retain all 272 predicted punctuation marks; readable SRT has 137 cues, max 7.94 s / 42 chars
• 224.427 s comparison sample remains aligned: 153 timestamped sentences retain all 152 punctuation marks

Tracks #3539. Keep the issue open until the reporter confirms the published artifact on the original workflow.

Signed-off-by: LauraGPT 18321252+LauraGPT@users.noreply.github.com