Cut down unit test code using destructor

[jwallwork23]

This PR tidies up the unit tests considerably in light of #297.

There's no longer any need to call torch_tensor_delete (except for the test of it). I also spotted a way to refactor the error raising so that we don't need all the clean_up subroutines.

In some cases, we can use the inbuilt pFUnit assertions, which is nice.

[jatkinson1000]

Yes, all makes sense to me. Thanks @jwallwork23
I like that the nullification is now at the same level as declaration of tensors/pointers and agree we should use the pFUnit preprocessor directives where we can.

Couple of questions around @assertEqual:

• What is the output of @assertEqual? Previously you had very explicit error messages relating to the test/routine/rank whereas I presume they are now more opaque/generic?
• I presume you checked that @assertEqual fails as expected when you break something? 😉

[jwallwork23]

* What is the output of `@assertEqual`? Previously you had very explicit error messages relating to the test/routine/rank whereas I presume they are now more opaque/generic?

So with the diff

diff --git a/test/unit/test_tensor_constructors.pf b/test/unit/test_tensor_constructors.pf
index 5e0e2c8..3b7d94a 100644
--- a/test/unit/test_tensor_constructors.pf
+++ b/test/unit/test_tensor_constructors.pf
@@ -93,7 +93,7 @@ contains
     call torch_tensor_delete(tensor)
 
     ! Check torch_tensor_delete does indeed free the memory
-    @assertFalse(c_associated(tensor%p))
+    @assertTrue(c_associated(tensor%p))
 
   end subroutine test_torch_tensor_empty

we get the failure

1: Failure
1:  in: 
1: test_tensor_constructors_suite.test_torch_tensor_empty[F][F]
1:   Location: 
1: [test_tensor_constructors.pf:96]
1: 
1:   
1: Failure
1:  in: 
1: test_tensor_constructors_suite.test_torch_tensor_empty[T][T]
1:   Location: 
1: [test_tensor_constructors.pf:96]

Personally, I think the line number is sufficient for reports of failed assertions.

* I presume you checked that `@assertEqual` fails as expected when you break something? 😉

Yep