fix(tests): improve subprocess error reporting in MCP tests

Replace subprocess.check_call() with subprocess.run() for pip operations.
This ensures stderr is captured and included in CalledProcessError exceptions,
making CI failures easier to debug.

- Use capture_output=True and text=True for better error visibility
- Applies to both pip install and pip uninstall operations

Co-authored-by: neelay-aign <190385684+neelay-aign@users.noreply.github.com>

Author: Copilot

tests/aignostics/utils/mcp_test.py

@@ -209,10 +209,11 @@ def install_dummy_mcp_plugin() -> Iterator[None]:
     import importlib
     import site
 
-    subprocess.check_call(
+    result = subprocess.run(
         [sys.executable, "-m", "pip", "install", "-e", str(DUMMY_PLUGIN_DIR)],
-        stdout=subprocess.DEVNULL,
-        stderr=subprocess.PIPE,
+        check=True,
+        capture_output=True,
+        text=True,
     )
 
     importlib.invalidate_caches()
@@ -221,10 +222,11 @@ def install_dummy_mcp_plugin() -> Iterator[None]:
 
     yield
 
-    subprocess.check_call(
+    result = subprocess.run(
         [sys.executable, "-m", "pip", "uninstall", "-y", "mcp-dummy-plugin"],
-        stdout=subprocess.DEVNULL,
-        stderr=subprocess.PIPE,
+        check=True,
+        capture_output=True,
+        text=True,
     )