Fix flaky test_big_file test by increasing client timeout (#1324)

Summary:
The test was failing with SocketTimeout because the pywatchman client in WatchmanInstance.start() used the default 1 second timeout, which is too short on resource-constrained systems (e.g., Windows CI running tests with 2GB/4GB files).

Changes:
- Increase client connection timeout from 1s to up to 10s
- Catch SocketTimeout in addition to SocketConnectError so the startup loop can retry on timeout

Pull Request resolved: #1324

Reviewed By: genevievehelsel

Differential Revision: D90087810

fbshipit-source-id: 4ad3c33229b0b2a2cff89bb4a820d2b85f433238

Author: ptarjan

watchman/integration/lib/WatchmanInstance.py

@@ -217,10 +217,16 @@ def start(self, extra_env=None) -> None:
         deadline = time.time() + self.start_timeout
         while time.time() < deadline:
             try:
-                client = pywatchman.client(sockpath=self.getSockPath())
+                # Use a reasonable timeout for individual connection attempts.
+                # This needs to be long enough to handle slow startup on
+                # resource-constrained systems (e.g., Windows CI with large files).
+                client = pywatchman.client(
+                    sockpath=self.getSockPath(),
+                    timeout=min(10.0, max(1.0, deadline - time.time())),
+                )
                 self.pid = client.query("get-pid")["pid"]
                 break
-            except pywatchman.SocketConnectError:
+            except (pywatchman.SocketConnectError, pywatchman.SocketTimeout):
                 t, val, tb = sys.exc_info()
                 time.sleep(0.1)
             finally: