feat(phase32): background task manager, interactive clarification modal, and diagnostics feedback extractor

[De-pitcher]

Phase 32: Background Tasks, Clarification Modal, and Diagnostics Feedback Extraction

Summary of Implementations

1. Background Task Execution & Process Lifecycle Management (manage_task /
   un_background):

   • Implemented agentcli/agent/tasks.py with \BackgroundTask\ and \TaskManager\ engine supporting
     un, \list, \status, \logs, \send_input, and \kill.
   • Bounded ring buffer log streaming (1000 lines default), non-blocking child output readers, Windows process group isolation, and automatic session cleanup on exit.
   • Implemented agentcli/subagents/task_manager.py registered as \SubAgentType.TASK_MANAGER.
   • Added /tasks\ slash command in CLI and Textual TUI dashboard.

2. Interactive Clarification & User Disambiguation (\�sk_user_question):

   • Implemented agentcli/subagents/clarification.py (\SubAgentType.CLARIFICATION).
   • Supports single-select and multi-select questionnaires with default recommendations and fallback for non-interactive / CI headless modes.

3. Compiler & Linter Diagnostics Feedback Extraction (\diagnostics_check):

   • Implemented agentcli/tools/diagnostics.py with \DiagnosticsParser\ and agentcli/subagents/diagnostics.py (\SubAgentType.DIAGNOSTICS).
   • Structured parsing for Pytest, Ruff/Flake8, Mypy, TypeScript (\ sc), ESLint, and Cargo/Rustc error spans with file, line, col, severity, and rule IDs.

4. Integration & Quality Gates:

   • Fully registered tools and subagents in \ToolRegistry, \AgentSession, and \ ools_schema.py.
   • 491/491 tests passing (100%).
   • 86.08% overall statement coverage (exceeding >= 85% requirement).
   • 0
     uff\ lint errors, 0 \mypy\ type errors across 72 source files.