Remove NCCL checkbox — auto-enable interconnect step when >1 GPU selected

NCCL all_reduce is always attempted when 2+ GPUs are selected; a failure
leaves InterconnectScore=0 (no bonus, no penalty) and OverallStatus
unaffected. Exposing the checkbox implied NCCL is optional and made a
failed run look like a deliberate skip.

- Remove benchmark-run-nccl checkbox and its change listener from pages.go
- Client sends run_nccl: selected.length > 1 (automatic)
- api.go default runNCCL=true is unchanged
- Selection note now mentions NCCL automatically for multi-GPU runs

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

audit/internal/webui/pages.go

@@ -1974,10 +1974,6 @@ func renderBenchmark(opts HandlerOptions) string {
         <input type="checkbox" id="benchmark-parallel-gpus">
         <span>Run all selected GPUs simultaneously (parallel mode, ignored in ramp-up)</span>
       </label>
-      <label class="benchmark-cb-row">
-        <input type="checkbox" id="benchmark-run-nccl" checked>
-        <span>Run multi-GPU interconnect step (NCCL) only on the selected GPUs</span>
-      </label>
       <p id="benchmark-selection-note" style="font-size:12px;color:var(--muted);margin:10px 0 14px">Select one GPU for single-card benchmarking or several GPUs for a constrained multi-GPU run.</p>
       <button id="benchmark-run-btn" class="btn btn-primary" onclick="runNvidiaBenchmark()" disabled>&#9654; Run Benchmark</button>
       <span id="benchmark-run-status" style="margin-left:10px;font-size:12px;color:var(--muted)"></span>
@@ -2033,7 +2029,6 @@ function benchmarkUpdateSelectionNote() {
   const selected = benchmarkSelectedGPUIndices();
   const btn = document.getElementById('benchmark-run-btn');
   const note = document.getElementById('benchmark-selection-note');
-  const nccl = document.getElementById('benchmark-run-nccl');
   if (!selected.length) {
     btn.disabled = true;
     note.textContent = 'Select at least one NVIDIA GPU to run the benchmark.';
@@ -2042,14 +2037,9 @@ function benchmarkUpdateSelectionNote() {
   btn.disabled = false;
   const rampUp = selected.length > 1 && !!document.getElementById('benchmark-ramp-up').checked;
   if (rampUp) {
-    note.textContent = 'Ramp-up: will spawn ' + selected.length + ' tasks (1 GPU → ' + selected.length + ' GPUs). NCCL runs on the final step only.';
+    note.textContent = 'Ramp-up: will spawn ' + selected.length + ' tasks (1 GPU → ' + selected.length + ' GPUs). NCCL interconnect runs on the final step.';
   } else {
-    note.textContent = 'Selected GPUs: ' + selected.join(', ') + '.';
-    if (nccl && nccl.checked && selected.length < 2) {
-      note.textContent += ' NCCL will be skipped because fewer than 2 GPUs are selected.';
-    } else if (nccl && nccl.checked) {
-      note.textContent += ' NCCL interconnect will use only these GPUs.';
-    }
+    note.textContent = 'Selected GPUs: ' + selected.join(', ') + '.' + (selected.length > 1 ? ' NCCL interconnect will be included.' : '');
   }
 }
 
@@ -2109,7 +2099,7 @@ function runNvidiaBenchmark() {
   const body = {
     profile: document.getElementById('benchmark-profile').value || 'standard',
     gpu_indices: selected,
-    run_nccl: !!document.getElementById('benchmark-run-nccl').checked,
+    run_nccl: selected.length > 1,
     parallel_gpus: parallelGPUs,
     ramp_up: rampUp,
     display_name: 'NVIDIA Benchmark'
@@ -2166,7 +2156,6 @@ function runNvidiaBenchmark() {
   });
 }
 
-document.getElementById('benchmark-run-nccl').addEventListener('change', benchmarkUpdateSelectionNote);
 benchmarkLoadGPUs();
 </script>`
 }