feat(dictation): transcribe after recording instead of live

Parakeet is an offline model trained on whole utterances, so re-decoding
the growing buffer to animate a live transcript cost O(n^2) work for a
result the final decode replaced. Sessions now decode once per committed
segment, and the composer shows a scrolling waveform of the mic level
instead of running text.

Long dictations split at a pause once past 60s (hard cap 90s) instead of
on a blind 15s timer, so cuts no longer land mid-word. Committed segments
decode while the user is still speaking: a 185s dictation returns 4.1s
after stop instead of 11.0s, with identical text (816 vs 817 words).

Also fixes two ways the stream manager could silently drop transcribed
audio. It now counts the commits it issued instead of trusting the
session's echoed events, so a commit still in flight when the client
finishes can no longer be left out of the final text. And segment
byte/peak accounting is reset where the commit is issued rather than when
the event arrives, which could mistake the tail of a dictation for
silence and clear it.

CHANGELOG.md

@@ -4,6 +4,7 @@ All notable changes to this project will be documented in this file.
 
 ## [Unreleased]
 
+- **Dictation:** speech is now transcribed after you stop talking, instead of being re-guessed word by word while you speak. The offline models OpenChamber runs are built to read a whole utterance at once, so the running transcript was consistently worse than the final one. While recording, the composer shows a live waveform of your voice and a timer, then Transcribing while the text is produced. Long recordings are split at pauses in your speech rather than on a timer, so a three-minute dictation still returns a few seconds after you stop, and words are no longer cut in half at the split.
 - Chat: if OpenCode restarts while a response is still running, the chat now stops with an interrupted state and a notification to continue instead of hanging silently (thanks to @sum117).
 
 ## [1.19.0] - 2026-08-19