fix(server): recover Prime turns after daemon reconnect

[rynfar]

A live Prime Agent turn could fail when only Pylon’s daemon transport reconnected, even though Prime kept the native execution alive. The recovered prompt request could also remain unresolved or reject after the worker had already admitted it.

This change keeps the existing adapter context and turn through a proven in-process reconnect. It accepts complete Prime replay metadata or an exact completed-message suffix, reconstructs missing response/tool/child state, gates descendant quiescence on the new connection generation, and fails closed when continuity cannot be proven. It never sends an admitted prompt again. Pylon process-restart adoption remains separate in #84.

Verification:

• 212 focused Prime adapter/runtime/MCP tests
• server typecheck
• targeted lint and formatting
• real Prime Agent 0.8.0 client-only transport fault during IPython tool work; the original prompt appeared exactly once and completed with PYLON_TRANSPORT_RECONNECT_OK

Closes #79

Model: GPT-5.6 Sol
Harness: Prime Agent

---

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.3 KiB	13.3 KiB	+7 B (+0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+7 B (+0.1%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.6 KiB	55.6 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	+54 B (+0.4%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+7 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.5 KiB	+47 B (+0.7%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.4 KiB	+88 B (+0.2%)	66.4 KiB	✅
Claude	Live turn messages	9	11	+2 (+22.2%)	21	✅

Baseline: 1baf147 · PR result: 89d3b9a · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.