feat(speech): read agent replies aloud

[n0tlu5]

Refs #844

Stacked on #853 (feat/speech-input). GitHub can't base a PR on a fork branch, so this targets main and the diff includes the earlier PRs. Please review only the commits below; I'll rebase once the earlier PR merges.

What

Read-aloud for agent replies: a speaker action on each reply, optional auto-read at turn end (active tab only), browser voices or an OpenAI-compatible speech endpoint, with code blocks and tables replaced by short spoken labels. Playback stops when you type, send, switch tabs or start dictation.

Commits to review

• feat(speech): add cloud speech synthesis backend
• feat(speech): extract speakable text from agent replies
• feat(speech): add read-aloud player
• feat(chat): add read-aloud action to agent replies
• feat(chat): auto-read agent replies at turn end

Checks

pnpm lint ., pnpm test, pnpm build; desktop cargo check, cargo clippy --all-targets --features test-utils -- -D warnings and cargo test --features test-utils; server cargo check, cargo test --lib and cargo clippy -D warnings; codeg-mcp cargo check and clippy. All pass at the tip of the stack, and the behaviour was exercised in web mode in Chromium. No new npm or cargo dependencies.