feat(core): Prepend vercel ai attributes with vercel.ai.X (#16908)

resolves https://linear.app/getsentry/issue/JS-663

Categorizing the attributes this way helps us differentiate between the
ai attributes added by vercel compared to the otel semantic conventions.

Author: AbhiPrasad

dev-packages/e2e-tests/test-applications/nextjs-15/tests/ai-test.test.ts

@@ -31,34 +31,34 @@ test('should create AI spans with correct attributes', async ({ page }) => {
 
   // First AI call - should have telemetry enabled and record inputs/outputs (sendDefaultPii: true)
   /* const firstPipelineSpan = aiPipelineSpans[0];
-  expect(firstPipelineSpan?.data?.['ai.model.id']).toBe('mock-model-id');
-  expect(firstPipelineSpan?.data?.['ai.model.provider']).toBe('mock-provider');
-  expect(firstPipelineSpan?.data?.['ai.prompt']).toContain('Where is the first span?');
+  expect(firstPipelineSpan?.data?.['vercel.ai.model.id']).toBe('mock-model-id');
+  expect(firstPipelineSpan?.data?.['vercel.ai.model.provider']).toBe('mock-provider');
+  expect(firstPipelineSpan?.data?.['vercel.ai.prompt']).toContain('Where is the first span?');
   expect(firstPipelineSpan?.data?.['gen_ai.response.text']).toBe('First span here!');
   expect(firstPipelineSpan?.data?.['gen_ai.usage.input_tokens']).toBe(10);
   expect(firstPipelineSpan?.data?.['gen_ai.usage.output_tokens']).toBe(20); */
 
   // Second AI call - explicitly enabled telemetry
   const secondPipelineSpan = aiPipelineSpans[0];
-  expect(secondPipelineSpan?.data?.['ai.prompt']).toContain('Where is the second span?');
+  expect(secondPipelineSpan?.data?.['vercel.ai.prompt']).toContain('Where is the second span?');
   expect(secondPipelineSpan?.data?.['gen_ai.response.text']).toContain('Second span here!');
 
   // Third AI call - with tool calls
   /*  const thirdPipelineSpan = aiPipelineSpans[2];
-  expect(thirdPipelineSpan?.data?.['ai.response.finishReason']).toBe('tool-calls');
+  expect(thirdPipelineSpan?.data?.['vercel.ai.response.finishReason']).toBe('tool-calls');
   expect(thirdPipelineSpan?.data?.['gen_ai.usage.input_tokens']).toBe(15);
   expect(thirdPipelineSpan?.data?.['gen_ai.usage.output_tokens']).toBe(25); */
 
   // Tool call span
   /*  const toolSpan = toolCallSpans[0];
-  expect(toolSpan?.data?.['ai.toolCall.name']).toBe('getWeather');
-  expect(toolSpan?.data?.['ai.toolCall.id']).toBe('call-1');
-  expect(toolSpan?.data?.['ai.toolCall.args']).toContain('San Francisco');
-  expect(toolSpan?.data?.['ai.toolCall.result']).toContain('Sunny, 72°F'); */
+  expect(toolSpan?.data?.['vercel.ai.toolCall.name']).toBe('getWeather');
+  expect(toolSpan?.data?.['vercel.ai.toolCall.id']).toBe('call-1');
+  expect(toolSpan?.data?.['vercel.ai.toolCall.args']).toContain('San Francisco');
+  expect(toolSpan?.data?.['vercel.ai.toolCall.result']).toContain('Sunny, 72°F'); */
 
   // Verify the fourth call was not captured (telemetry disabled)
   const promptsInSpans = spans
-    .map(span => span.data?.['ai.prompt'])
+    .map(span => span.data?.['vercel.ai.prompt'])
     .filter((prompt): prompt is string => prompt !== undefined);
   const hasDisabledPrompt = promptsInSpans.some(prompt => prompt.includes('Where is the third span?'));
   expect(hasDisabledPrompt).toBe(false);