feat(phase35): Token Budget Governor & Adaptive Context Compression

[De-pitcher]

Phase 35: Token Budget Governor & Adaptive Context Compression

Summary of Changes

1. TokenBudgetGovernor (\�gentcli/memory/governor.py):
   • Real-time token and USD tracking across models, prompt tokens, completion tokens, cached tokens, and subagent types.
   • Dynamic spend velocity calculations ($/hr, \ okens/min).
   • Soft budget warning (80% threshold) and hard ceiling enforcement with detailed usage breakdowns and formatted summary tables.
2. AdaptiveContextCompressor (\�gentcli/memory/adaptive_compressor.py):
   • Progressive, multi-tier context compression (ANSI escape code removal, consecutive line deduplication, tool execution payload truncation with hash digests, assistant message trimming).
   • Compression telemetry and metrics reporting.
3. Session Integration (\�gentcli/session.py):
   • Integrated \TokenBudgetGovernor\ into
     ecord_cost\ and \is_budget_exceeded.
   • Wired \AdaptiveContextCompressor\ into _trim_history.
4. CLI & TUI Commands:
   • Enhanced /budget\ with [set | max-tokens | tier <low|med|high> | reset | status].
   • Formatted /cost\ and /tokens\ with governor statistics and breakdown.
   • Autocomplete support in \SlashAndFileCompleter\ and command dispatch in \TUIApp.
5. Comprehensive Test Suite:
   • Added \ ests/test_phase35_token_governor.py\ with 11 focused test cases covering all components.

Quality Verification

• Ruff: 0 errors
• Mypy: 0 errors across 127 files
• Tests: 100% passing across Phase 35, Session, CLI, Worktrees, and TUI test suites.