ci: broaden GPU runner family to g5.2xlarge+g6.2xlarge

Provisioning only g5.2xlarge leaves the GPU tests exposed to spot-capacity
exhaustion, surfacing as RunsOn LaunchError "unable to schedule this runner".
g6.2xlarge (L4, 24 GB) matches the g5 A10G VRAM class, so RunsOn can fall back
to whichever GPU capacity is available.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: pinin4fjords

.github/workflows/nf-test-gpu.yml

@@ -58,7 +58,7 @@ jobs:
 
   nf-test-gpu:
     needs: [get-shards]
-    runs-on: "runs-on=${{ github.run_id }}/family=g5.2xlarge/image=ubuntu24-gpu-x64"
+    runs-on: "runs-on=${{ github.run_id }}/family=g5.2xlarge+g6.2xlarge/image=ubuntu24-gpu-x64"
     name: "GPU Test | ${{ matrix.profile }} | ${{ matrix.shard }} | ${{ matrix.NXF_VER }} | ${{ matrix.filters }}"
     env:
       NXF_VER: ${{ matrix.NXF_VER }}