Proposal: Post-Debate Reasoning Analysis – Initial Scaffolding

## Context
DebateAI currently provides debate outcomes, judging, and rankings, but users have limited visibility into *why* specific arguments fail at a reasoning level.
Following recent Discord discussions, this issue proposes an initial, minimal step toward adding a post-debate reasoning and learning layer to DebateAI.

## Problem
While debates are structured (opening / cross-exam / closing), there is currently no logic-level analysis that:
- Identifies recurring reasoning mistakes
- Explains why certain argument patterns are weak
- Supports structured post-debate learning over time

## Proposed Initial Scope (Foundational Only)
This issue intentionally focuses on scaffolding, not features:
- Introduce a dedicated post-debate `analysis` module
- Define interfaces and data structures for future reasoning analysis
- Ensure the system runs only after debates conclude
- Keep the module optional and non-invasive

## Explicit Non-Goals (for this phase)
- No real-time debate intervention
- No judging, ranking, or Elo changes
- No AI opponent modifications
- No fallacy detection logic yet

## Outcome
If this structure is acceptable, follow-up PRs can incrementally add:
- Fallacy signal heuristics
- Explainable coaching
- Reasoning-aware history

Feedback on scope and structure is welcome before proceeding.