cost dialog: enrich with session stats, per-model percentages, and formatting fixes

Enhance the /cost dialog with richer session insights:

- Show total token count (input + output combined) in the Total section
- Display session duration and message count in the header
- Show reasoning tokens when the model reports them (e.g. o1)
- Add average cost per 1K tokens and average cost per message metrics
- Show cost percentage breakdown per model and per message
- Display cache hit rate and per-entry cached token counts
- Include model name in the plain-text clipboard copy output

Code quality improvements:

- Extract shared stat computation (totalStats, actualMessageCount,
  plainTextLine) to eliminate duplication between styled and plain-text
  renderers
- Fix slice aliasing bug in applyScrolling (use slices.Clone)
- Fix formatCost/formatCostPadded silently swallowing negative values
- Fix formatDuration producing garbled output for negative durations

Assisted-By: docker-agent

Author: dgageot