addon.node : VAD-aligned (faster-whisper-like) timeline with real gaps

With VAD enabled, whisper.cpp concatenates all detected speech into a
single stream, so addon.node returned a gap-less timeline where every
segment end equals the next segment start. Add an alignment layer that
puts timestamps back on the original timeline with real silence gaps,
controlled by three new params:

  align_mode       "hybrid" (default) | "run" | "word" | "legacy"
  vad_merge_gap_ms adjacent VAD segments whose silence gap is <= this (ms)
                   merge into one run; a larger gap becomes a real gap
                   (default 2000; negative disables aligned mode)
  word_gap_ms      word/hybrid: start a new segment when the gap between
                   two consecutive words exceeds this (default 500)

- hybrid (C): VAD-grouped runs, each run sliced into its own buffer and
  re-segmented by word-level gaps with every segment end clamped to its
  last word. Best boundary accuracy; default.
- run (A): per-run decode emitting whisper's own segments; gaps between runs.
- word (B): single decode pass + word-gap re-segmentation; uses core VAD
  via the new whisper_full_get_token_t0/t1 mapping when a VAD model is given.
- legacy: original continuous single pass.

Each run is decoded from a physically sliced buffer rather than via
offset_ms/duration_ms (which only bound the outer seek loop and let
neighbouring speech bleed into short runs); slice-relative timestamps are
shifted back with a per-run base offset. Progress is rescaled across runs
so the JS callback still sees a single monotonic 0..100.

Author: linxiaodong