fix(kernel,server,tui,inference): generic resource gauge + F-002/F-003/F-004

- generic SystemResourceProbe: owner-agnostic /proc/meminfo system RAM
  overlaid on the vendor GPU probe, wired on both managed and static paths
  so the VRAM/GPU/RAM gauge renders with an external llama-server (was dormant)
- F-002: classify provider 4xx (e.g. llama.cpp 400) as terminal, not retryable
  (except 408/429); stops retrying deterministic request failures to exhaustion
- F-003: FileSystemPromptLoader expands leading ~ / ~/ to user home
- F-004: map InferenceFailedEvent + RetryAttemptedEvent to new
  ServerMessage.InferenceFailed / RetryAttempted, decoded+rendered in tui-go;
  ArtifactContentStoredEvent explicitly mapped to null (internal, no warn)
- tests: tilde expansion, SystemResourceProbe (static/managed/fail-soft)

apps/tui-go/internal/app/server.go

@@ -3,6 +3,7 @@ package app
 import (
 	"crypto/rand"
 	"encoding/hex"
+	"fmt"
 	"strings"
 	"time"
 
@@ -172,6 +173,16 @@ func (m *Model) applyServer(msg protocol.ServerMessage) {
 			s.Active = false
 			s.addEvent(nowMillis(), "InferenceTimedOut", msg.StageID)
 		}
+	case protocol.TypeInferenceFailed:
+		if s := m.session(msg.SessionID); s != nil {
+			s.Active = false
+			s.addEvent(nowMillis(), "InferenceFailed", msg.StageID+": "+msg.Reason)
+		}
+	case protocol.TypeInferenceRetry:
+		if s := m.session(msg.SessionID); s != nil {
+			s.addEvent(nowMillis(), "RetryAttempted",
+				fmt.Sprintf("%s (%d/%d): %s", msg.StageID, msg.AttemptNumber, msg.MaxAttempts, msg.FailureReason))
+		}
 	case protocol.TypeToolStarted:
 		if s := m.session(msg.SessionID); s != nil {
 			s.Active = true
@@ -274,6 +285,8 @@ func (m *Model) applyServer(msg protocol.ServerMessage) {
 		m.gpuTotalMB = msg.GpuMemoryTotalMb
 		m.gpuUtil = msg.GpuUtilizationPct
 		m.ramMB = msg.ProcessRssMb
+		m.sysRamUsedMB = msg.SystemRamUsedMb
+		m.sysRamTotalMB = msg.SystemRamTotalMb
 	case protocol.TypeArtifactList:
 		m.artifacts = msg.Artifacts
 		m.artifactsFor = msg.SessionID