Merge debug/prod into single ISO build, fix NVIDIA module loading

## ISO build consolidation
- Remove separate debug/prod split: overlay-debug/, build-debug.sh,
  mkimg.bee_debug.sh, genapkovl-bee_debug.sh all deleted
- Single overlay: iso/overlay/ (was overlay-debug content)
- Single build script: build.sh (SSH, TUI, NVIDIA, vendor tools, bee-release)
- Single mkimage profile: bee (with dropbear, dialog, strace, gcompat, etc.)

## NVIDIA fixes
- Modules now stored at /usr/local/lib/nvidia/ instead of
  /lib/modules/<kver>/extra/nvidia/ — modloop squashfs mounts over that
  path at boot making overlay content there inaccessible
- bee-nvidia init: load via insmod (absolute path), not modprobe
- bee-nvidia init: create libnvidia-ml.so.1/libcuda.so.1 symlinks in /usr/lib/
- build-nvidia-module.sh: always install linux-lts-dev (not conditional) —
  stale 6.6.x headers caused wrong-kernel modules that never loaded at runtime
- build-nvidia-module.sh: create soname symlinks in cache
- KERNEL_VERSION in VERSIONS updated 6.6 → 6.12
- gcompat added to ISO packages (nvidia-smi is a glibc binary on musl Alpine)

## Service ordering
- bee-audit: add `after bee-nvidia` so NVIDIA enrichment always succeeds

## New tooling
- iso/builder/smoketest.sh: SSH smoke test for post-boot ISO validation
- iso/builder/build-gpu-burn.sh: builds gpu_burn vendor binary (CUDA 12.8+)
- vendor/gpu_burn included automatically if placed in iso/vendor/

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

iso/overlay/etc/init.d/bee-nvidia

@@ -2,6 +2,8 @@
 
 description="Bee: load NVIDIA kernel modules"
 
+NVIDIA_KO_DIR="/usr/local/lib/nvidia"
+
 depend() {
     need localmount
     before bee-audit
@@ -9,23 +11,39 @@ depend() {
 
 start() {
     ebegin "Loading NVIDIA modules"
-    kver="$(uname -r)"
-    einfo "kernel: ${kver}"
-    if [ -d "/lib/modules/${kver}/extra/nvidia" ]; then
-        einfo "module dir: /lib/modules/${kver}/extra/nvidia"
-        ls "/lib/modules/${kver}/extra/nvidia"/*.ko 2>/dev/null | sed 's/^/  /' || true
-    else
-        ewarn "module dir missing: /lib/modules/${kver}/extra/nvidia"
+    einfo "kernel: $(uname -r)"
+
+    if [ ! -d "$NVIDIA_KO_DIR" ]; then
+        ewarn "NVIDIA module dir missing: $NVIDIA_KO_DIR"
+        eend 1
+        return 1
     fi
 
-    depmod -a 2>/dev/null || true
+    einfo "module dir: $NVIDIA_KO_DIR"
+    ls "$NVIDIA_KO_DIR"/*.ko 2>/dev/null | sed 's/^/  /' || true
 
+    # Create libnvidia-ml soname symlinks needed by nvidia-smi (glibc binary on Alpine/musl)
+    for lib in libnvidia-ml libcuda; do
+        versioned=$(ls /usr/lib/${lib}.so.[0-9]* 2>/dev/null | head -1)
+        [ -n "$versioned" ] || continue
+        base=$(basename "$versioned")
+        ln -sf "$base" "/usr/lib/${lib}.so.1" 2>/dev/null || true
+        ln -sf "${lib}.so.1" "/usr/lib/${lib}.so" 2>/dev/null || true
+    done
+
+    # Load modules via insmod (bypasses modules.dep — modloop squashfs is read-only)
     for mod in nvidia nvidia-modeset nvidia-uvm; do
-        if modprobe "$mod" 2>/dev/null; then
-            einfo "loaded: $mod"
+        ko="$NVIDIA_KO_DIR/${mod}.ko"
+        [ -f "$ko" ] || ko="$NVIDIA_KO_DIR/${mod//-/_}.ko"
+        if [ -f "$ko" ]; then
+            if insmod "$ko" 2>/dev/null; then
+                einfo "loaded: $mod"
+            else
+                ewarn "failed to load: $mod"
+                dmesg | tail -n 5 | sed 's/^/  dmesg: /' || true
+            fi
         else
-            ewarn "failed to load: $mod"
-            dmesg | tail -n 5 | sed 's/^/  dmesg: /' || true
+            ewarn "not found: $ko"
         fi
     done