[Tests] Fix extension in test case for apple targets. (#390)

**Context:** Apple uses dylib

**Description of the Change:** Use dylib

doc/changelog.md

@@ -19,6 +19,7 @@
 * Support for dynamically-shaped arrays has been added.
   [(#366)](https://github.com/PennyLaneAI/catalyst/pull/366)
   [(#386)](https://github.com/PennyLaneAI/catalyst/pull/385)
+  [(#390)](https://github.com/PennyLaneAI/catalyst/pull/390)
 
   Catalyst now accepts tensors whose dimensions are not known at compile time.
   Standard tensor initialisation functions `jax.numpy.ones`, `jnp.zeros`, and

frontend/test/pytest/test_compiler.py

@@ -266,6 +266,7 @@ def test_compiler_driver_with_flags(self):
     def test_compiler_from_textual_ir(self):
         """Test the textual IR compilation."""
         full_path = get_lib_path("runtime", "RUNTIME_LIB_DIR")
+        extension = ".so" if platform.system() == "Linux" else ".dylib"
 
         # pylint: disable=line-too-long
         ir = (
@@ -281,7 +282,9 @@ def test_compiler_from_textual_ir(self):
     quantum.device ["rtd_lib", """
             + r'"'
             + full_path
-            + r"""/librtd_lightning.so"]
+            + r"""/librtd_lightning"""
+            + extension
+            + """"]
     %0 = stablehlo.constant dense<4> : tensor<i64>
     %1 = quantum.alloc( 4) : !quantum.reg
     %2 = stablehlo.constant dense<0> : tensor<i64>