Add insertBits edge case regression test (#4664)

This reproduces a GPU hang on Mac for the Dawn implementation (see
crbug.com/506180954).

src/webgpu/listing_meta.json

@@ -1491,6 +1491,7 @@
   "webgpu:shader,execution,expression,call,builtin,fwidth:f32:*": { "subcaseMS": 29.807 },
   "webgpu:shader,execution,expression,call,builtin,fwidthCoarse:f32:*": { "subcaseMS": 17.110 },
   "webgpu:shader,execution,expression,call,builtin,fwidthFine:f32:*": { "subcaseMS": 16.906 },
+  "webgpu:shader,execution,expression,call,builtin,insertBits:array_index:*": { "subcaseMS": 20.682 },
   "webgpu:shader,execution,expression,call,builtin,insertBits:integer:*": { "subcaseMS": 9.569 },
   "webgpu:shader,execution,expression,call,builtin,inversesqrt:abstract_float:*": { "subcaseMS": 19408.045 },
   "webgpu:shader,execution,expression,call,builtin,inversesqrt:f16:*": { "subcaseMS": 21.411 },

src/webgpu/shader/execution/expression/call/builtin/insertBits.spec.ts

@@ -378,3 +378,46 @@ g.test('integer')
 
     await run(t, builtin('insertBits'), [T, T, Type.u32, Type.u32], T, cfg, cases);
   });
+
+g.test('array_index')
+  .desc('Test insertBits offset edge case where the result is used as an array index.')
+  .fn(t => {
+    const wgsl = `
+      @group(0) @binding(0) var<storage, read_write> res: array<u32, 8>;
+
+      @compute @workgroup_size(1)
+      fn main() {
+        let count = 0u;
+        res[insertBits(5u, 1u, 32u, count)] = 42u;
+      }
+    `;
+
+    const pipeline = t.device.createComputePipeline({
+      layout: 'auto',
+      compute: {
+        module: t.device.createShaderModule({ code: wgsl }),
+        entryPoint: 'main',
+      },
+    });
+
+    const resBuffer = t.createBufferTracked({
+      size: 8 * 4,
+      usage: GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC,
+    });
+
+    const group = t.device.createBindGroup({
+      layout: pipeline.getBindGroupLayout(0),
+      entries: [{ binding: 0, resource: { buffer: resBuffer } }],
+    });
+
+    const encoder = t.device.createCommandEncoder();
+    const pass = encoder.beginComputePass();
+    pass.setPipeline(pipeline);
+    pass.setBindGroup(0, group);
+    pass.dispatchWorkgroups(1);
+    pass.end();
+
+    t.queue.submit([encoder.finish()]);
+
+    t.expectGPUBufferValuesEqual(resBuffer, new Uint32Array([0, 0, 0, 0, 0, 42, 0, 0]));
+  });