feat(gpu): show VRAM in use as a bar with a percentage

Sum drm-resident-vram0 from DRM fdinfo, the same per-client counters
gputop reads, counting each drm-client-id once since a client repeats
its totals on every fd it holds. One grep over /proc/*/fdinfo, about
10ms warm, sampled through new_gate every ~2s.

The 12 GiB total is hardcoded: xe exposes no VRAM capacity in sysfs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

1 files changed, 11 insertions, 5 deletions

diff --git a/README.md b/README.md
index 3de25f7..b137121 100644
--- a/README.md
+++ b/README.md
@@ -68,7 +68,8 @@ Seven cards cover the machine itself:
 - **System**: CPU load as a per-core equaliser, the CPU and motherboard model
   as two rows under the header, RAM used/total, and every temperature (CPU
   package, CCD die, NVMe and two board sensors).
-- **GPU**: Arc B580 package and VRAM temperatures, and fan speed.
+- **GPU**: Arc B580 package and VRAM temperatures, VRAM in use as a bar, and
+  fan speed.
 - **Disks**: `/`, `/home`, `/data` and one NFS mount per server, as usage
   rings, with free space and total size printed under each ring. Two servers
   are configured, each with the same export mounted twice; the card shows one
@@ -488,10 +489,15 @@ card that computed free space as `size - used` would overstate it by exactly
 that reserve.
 
 **The Arc B580 reports no utilisation through any stable interface.** The `xe`
-driver exposes no `gpu_busy_percent` (unlike `amdgpu`), `intel_gpu_top` refuses
-the device, and `gputop` prints per-process rows with ANSI escapes, which is not
-something to build a widget on. The GPU card therefore has no load bar; it shows
-what the hardware actually reports rather than inventing a number.
+driver exposes no `gpu_busy_percent` (unlike `amdgpu`) and `intel_gpu_top`
+refuses the device. The GPU card therefore has no load bar; it shows what the
+hardware actually reports rather than inventing a number.
+
+**VRAM in use comes from DRM fdinfo**, the per-client counters `gputop` reads:
+`drm-resident-vram0` in `/proc/<pid>/fdinfo/<fd>`, summed once per
+`drm-client-id`. Only processes readable by conky's user are seen, so a GPU
+client running as another user is not counted. The card's 12 GiB total is
+hardcoded: xe puts no VRAM capacity in sysfs.
 
 **`/data` has no temperature.** It sits on a spinning disk with no hwmon chip;
 reading its SMART temperature would need `smartctl` as root and would spin up a