Add manual speaker controls

[wassgha]

Summary

Speaker labels in the transcript are no longer static Speaker N text. Users can rename, reassign, add, move, merge, and remove them:

• Rename — click a label (or use the ⋯ menu) and type a new name; updates every occurrence
• Change speaker — pick another speaker (or create one) for a whole turn, or for a word selection via the toolbar Speaker button / @
• Add speaker — create a new named speaker while changing a turn or selection
• Move label — drag the grip on a turn header to slide the boundary with the previous speaker
• Replace in project — merge all of one speaker into another
• Remove from project — drop a speaker; their turns join the speaker above in the script

Implementation notes

• New SpeakerInfo { id, name } list lives in editor state and undo snapshots
• Pure helpers in lib/speakers.ts (covered by tests/speakers-test.ts)
• Names round-trip through IndexedDB projects and SRT / VTT / JSON / TXT / MD export
• JSON export now includes a speakers array for named round-trips
• Selection freeze shared with Correct so the Speaker picker can take focus without clearing the selection

Testing

• npx tsx tests/speakers-test.ts
• npx tsx tests/parse-transcript-test.ts
• npx tsx tests/serialize-transcript-test.ts
• npx tsc --noEmit
• Playwright UI smoke: rename, create from selection, move boundary, remove