ux: /compact should show before/after context usage

As an extension of #1313 , we should show before and after usage stats, so that we can understand the impact of both compaction and user rules files on the size of the requests they are sending to the model.

eg.
Current context window
Current context window (190k of 200k tokens used)
███████████████████████████████████████████████████████████████████████████████|█ 98.60%

█ Context files: ~30850 tokens (15.42%)
█ Q responses: ~32360 tokens (0.18%)
█ Your prompts: ~492301 tokens (83.00%)

Compacting.....
✔ Conversation history has been compacted successfully!

• The assistant has access to all previous tool executions, code analysis, and discussion details
• The assistant will reference specific information from the summary when relevant

New context window (31200 of 200k tokens used)
████████████|████████████████████████████████████████████████████████████████████ 15.60%

█ Context files: ~30850 tokens (15.42%)
█ Q responses: ~360 tokens (0.18%)
█ Your prompts: ~0 tokens (0.00%

════════════════════════════════════════════════════════════════════════════════
CONVERSATION SUMMARY
════════════════════════════════════════════════════════════════════════════════

lots of text
This summary is stored in memory and available to the assistant.
It contains all important details from previous interactions.
════════════════════════════════════════════════════════════════════════════════