fix(nvidia): surface Xid 79/154 GPU bus-fall-off as a physical-reboot-required signal

nvidia-bug-report.sh appends .gz to --output-file when gzip is available,
which silently produced empty nvidia-bug-report.txt in support bundles
(cat looked for the uncompressed name that never existed).

Also: a GPU that falls off the PCIe/NVLink bus (Xid 79) or gets flagged
for Node Reboot Required (Xid 154) mid-SAT-run left every downstream test
failing with generic, unrelated-looking errors (CUDA "unknown error",
"unable to determine device handle") with no indication the GPU needed a
physical power-cycle to recover. Detect these codes from SAT run logs and
surface a plain-English "physical reboot required" message in the task's
failure detail, the persisted component-status DB, a dashboard banner on
the Hardware Summary card, and topology diagram GPU-node severity.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

audit/internal/collector/xid.go

@@ -11,6 +11,10 @@ var xidCodeParenRE = regexp.MustCompile(`(?i)Xid\s*\([^)]*\)\s*:?\s*(\d+)`)
 // xidCodeColonRE handles the older "Xid: 64, ..." form with no BDF parens.
 var xidCodeColonRE = regexp.MustCompile(`(?i)\bXid\s*:\s*(\d+)`)
 
+// xidCodeBareRE handles dcgmi diag's own report wording, e.g.
+// "Detected XID 79 for GPU 1" — no colon or parens around the code.
+var xidCodeBareRE = regexp.MustCompile(`(?i)\bXid\s+(\d+)\b`)
+
 // xidCodeSeverity maps NVIDIA Xid codes relevant to GPU HBM/ECC health to a
 // severity, refining the generic "nvidia-xid" kernel-log pattern's default
 // "warning". Xid 64 is the same InfoROM row-remap-write failure surfaced by
@@ -23,7 +27,9 @@ var xidCodeColonRE = regexp.MustCompile(`(?i)\bXid\s*:\s*(\d+)`)
 var xidCodeSeverity = map[string]string{
 	"48":  "critical",
 	"64":  "critical",
+	"79":  "critical",
 	"95":  "critical",
+	"154": "critical",
 	"63":  "warning",
 	"94":  "warning",
 	"160": "warning",
@@ -42,6 +48,29 @@ func XidSeverity(line string) (severity string, ok bool) {
 	return sev, ok
 }
 
+// xidHardwareFaultMessages maps NVIDIA Xid codes that mean the GPU has left
+// the bus and cannot self-recover to a plain-English explanation, so an
+// operator sees "physical reboot required" directly instead of having to
+// look up what "Xid 79" means in NVIDIA's docs.
+// Source: NVIDIA Xid error docs (79 = GPU has fallen off the bus, 154 = GPU
+// recovery action escalated to Node Reboot Required).
+var xidHardwareFaultMessages = map[string]string{
+	"79":  "GPU has fallen off the PCIe bus (Xid 79) — hardware fault, will not recover without a physical reboot/power-cycle",
+	"154": "NVIDIA driver flagged this GPU for a required node reboot (Xid 154: recovery action = Node Reboot Required)",
+}
+
+// XidHardwareFaultMessage returns a plain-English explanation for a log line
+// carrying an NVIDIA Xid code that requires a physical reboot to clear, if
+// the code is recognized. ok is false otherwise.
+func XidHardwareFaultMessage(line string) (message string, ok bool) {
+	code, ok := extractXidCode(line)
+	if !ok {
+		return "", false
+	}
+	msg, ok := xidHardwareFaultMessages[code]
+	return msg, ok
+}
+
 func extractXidCode(line string) (string, bool) {
 	if m := xidCodeParenRE.FindStringSubmatch(line); m != nil {
 		return m[1], true
@@ -49,5 +78,8 @@ func extractXidCode(line string) (string, bool) {
 	if m := xidCodeColonRE.FindStringSubmatch(line); m != nil {
 		return m[1], true
 	}
+	if m := xidCodeBareRE.FindStringSubmatch(line); m != nil {
+		return m[1], true
+	}
 	return "", false
 }