fix: limit parallel browser instances and reject concurrent persistent sessions

- Persistent/existing mode: throw error when instance is already in-use,
  instructing the model to run browser tasks sequentially. This prevents
  Chrome profile lock conflicts from concurrent --userDataDir usage.
- Isolated mode: allow parallel instances up to MAX_PARALLEL_INSTANCES (5).
  Throw when the limit is exceeded to prevent resource exhaustion.
- Update integration test to use isolated mode for concurrent browser agents.
- Add 3 new unit tests for persistent/existing rejection and max limit.

Author: gsquared94

integration-tests/browser-agent.test.ts

@@ -308,7 +308,7 @@ describe.skipIf(!chromeAvailable)('browser-agent', () => {
     await run.expectText('successfully written', 15000);
   });
 
-  it('should handle concurrent browser agents with persistent session mode', async () => {
+  it('should handle concurrent browser agents with isolated session mode', async () => {
     rig.setup('browser-concurrent', {
       fakeResponsesPath: join(__dirname, 'browser-agent.concurrent.responses'),
       settings: {
@@ -320,9 +320,10 @@ describe.skipIf(!chromeAvailable)('browser-agent', () => {
           },
           browser: {
             headless: true,
-            // Persistent mode is the default — the regression scenario.
-            // Two concurrent calls should each get a separate browser.
-            sessionMode: 'persistent',
+            // Isolated mode supports concurrent browser agents.
+            // Persistent/existing modes reject concurrent calls to prevent
+            // Chrome profile lock conflicts.
+            sessionMode: 'isolated',
           },
         },
       },