fix(core): Address review nits on btoa base64 encoder

- Soften the JSDoc claim about JSC: not every JSC build exposes `btoa`,
  only some do.
- Add a test covering the JS-encoder fallback path by stashing
  `globalThis.btoa` so the module sees no native implementation.

Author: alwx

packages/core/src/js/utils/base64.ts

@@ -6,10 +6,10 @@ const CHUNK_SIZE = 0x8000;
 
 /**
  * Encodes a byte array to base64. Uses the runtime's native `btoa` when
- * available (Hermes and modern JSC both expose it), which is significantly
- * faster than the pure-JS encoder for the envelope payloads going through
- * `RNSentry.captureEnvelope`. Falls back to the bundled JS encoder when
- * `btoa` is missing (e.g. older JS engines).
+ * available — Hermes exposes it natively, and some JSC builds do too — which
+ * is significantly faster than the pure-JS encoder for the envelope payloads
+ * going through `RNSentry.captureEnvelope`. Falls back to the bundled JS
+ * encoder when `btoa` is missing.
  */
 export function encodeToBase64(bytes: Uint8Array): string {
   const nativeBtoa = (globalThis as { btoa?: (input: string) => string }).btoa;

packages/core/test/utils/base64.test.ts

@@ -15,4 +15,17 @@ describe('encodeToBase64', () => {
     }
     expect(encodeToBase64(bytes)).toBe(Buffer.from(bytes).toString('base64'));
   });
+
+  test('falls back to the JS encoder when `btoa` is unavailable', () => {
+    const originalBtoa = (globalThis as { btoa?: unknown }).btoa;
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    delete (globalThis as any).btoa;
+    try {
+      // "sentry" => "c2VudHJ5"
+      expect(encodeToBase64(new Uint8Array([0x73, 0x65, 0x6e, 0x74, 0x72, 0x79]))).toBe('c2VudHJ5');
+    } finally {
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      (globalThis as any).btoa = originalBtoa;
+    }
+  });
 });