feat: Session Replay and Bridge Capability Mapping from Recorded User Flows

[Delqhi]

Summary

OpenSIN-Bridge should not only record sessions but also map them to what the bridge can already automate and where the gaps are.

Problem

We currently lack a feedback loop that answers:

• which recorded steps are already bridge-native
• which need new tools
• which need agent-side orchestration
• which fail because of shadow DOM, iframe, auth, or anti-bot constraints

Goal

Build a replay/mapping layer that:

1. replays candidate steps in dry-run / review mode
2. marks each step as bridge_supported, bridge_partial, or bridge_missing
3. generates implementation hints for missing capabilities

Suggested Output

• per-step support classification
• required bridge tools
• required new issues / missing tool suggestions
• replay evidence (DOM diff / screenshot diff / network diff)

Repo Evidence

• Existing observe, snapshot, page_diff, click_ref, type_ref, vision_* tools
• Existing recording/video + request log fragments
• Existing bridge/server split suitable for capability mapping

Acceptance Criteria

• Recorded session can be replayed in review mode
• Each step gets a capability mapping status
• Missing bridge capabilities are surfaced as structured gaps
• Output is suitable for agent planning / future GitHub issue generation