fix(transport): cap LineDiffer input size — oversized script diffs degrade to summary instead of O((N+M)^2) memory

Claude-Session: https://claude.ai/code/session_01MtdgwpEeCUn6cUA5f1LMPj

tests/ZB.MOM.WW.ScadaBridge.Transport.Tests/Import/LineDifferTests.cs

@@ -1,3 +1,5 @@
+using System.Diagnostics;
+
 using ZB.MOM.WW.ScadaBridge.Transport.Import;
 
 namespace ZB.MOM.WW.ScadaBridge.Transport.Tests.Import;
@@ -189,4 +191,29 @@ public sealed class LineDifferTests
         Assert.Equal(0, result.AddedCount);
         Assert.Equal(0, result.RemovedCount);
     }
+
+    [Fact]
+    public void OversizedInputs_ZeroCommonLines_DegradesToSummary_WithoutRunningMyers()
+    {
+        // Two 10,000-line inputs with zero common lines. Running the Myers trace here would
+        // record ~20,000 V-snapshots of ~160 KB each (multi-GB transient allocation), so the
+        // input-size guard must short-circuit to a summary-only result long before that.
+        var oldText = string.Join("\n", Enumerable.Range(1, 10_000).Select(i => $"old-line-{i}"));
+        var newText = string.Join("\n", Enumerable.Range(1, 10_000).Select(i => $"new-line-{i}"));
+
+        var sw = Stopwatch.StartNew();
+        var result = LineDiffer.Diff(oldText, newText);
+        sw.Stop();
+
+        // Summary-only shape: same as the output cap — Truncated, full totals, no hunks.
+        Assert.True(result.Truncated);
+        Assert.Empty(result.Lines);
+        Assert.Equal(10_000, result.AddedCount);
+        Assert.Equal(10_000, result.RemovedCount);
+
+        // Must return within a small time budget; pathological slowness is a failure.
+        Assert.True(
+            sw.Elapsed < TimeSpan.FromSeconds(2),
+            $"Diff of oversized inputs took {sw.Elapsed.TotalSeconds:F1}s — the input-size guard did not short-circuit.");
+    }
 }