feat(usage): aggregate token usage and expose in CLI view and across harnesses

[Adityakk9031]

Most harnesses track token usage in some form, but txcript previously dropped it during conversions, and xcript view didn't show any token statistics.

This PR wires up token tracking from the transcript model all the way through harnesses and into the terminal view.

What changed:

• Usage math & aggregation: Implemented Default, otal_tokens(), is_zero(), along with Add, AddAssign, and Sum for Usage. Added ranscript.total_usage() to sum token footprints across assistant turns.
• Harness support:
  • x: parses otal_input_tokens and otal_output_tokens from history_turn_committed payloads, falling back to usage-v2.json / session.json. Re-synthesizes accurate usage stats on export.
  • cursor_desktop: pulls token metrics from both okenCount and okenUsage schemas (prompt, completion, cache read/write).
  • hermes: reads usage columns from SQLite/export files and preserves structured usage objects during export.
• Terminal view ( xcript view):
  • Displays total session tokens in the header: \Tokens: ( in, out[, cached])\
  • Shows per-turn tokens in assistant message headers: \Message #N · Assistant · tokens\
• Tests & docs:
  • Added test suite in ests/integration/usage.rs testing arithmetic, aggregation, and lossless export roundtrips.
  • Updated docs/usage.md.

All tests pass and clippy runs clean.