Fix HAL MPU test on Cortex M7 CPUs

[multiplemonomials]

Summary of changes

Did some checking into finally fixing this annoying test failure that has been happening for years. The hal-mpu stack fault test has been dying on most MCUs 100% of the time, and it wasn't super obvious why.

When debugging the test, I noticed that it was dying because it was executing a garbage instruction (NOT because it was executing from RAM -- that's an intended part of the test). Why was this happening? Well, stepping through the disassembly of mpu_fault_test_stack(), it looked like the compiler had reordered the instructions, and was executing line 141 before line 147! So it was executing the RAM instruction before said instruction had been initialized with a value.

Why was that happening? Well, I think that the cast that this test does breaks aliasing rules in C++ -- it's casting an integer into a function pointer, and the optimizer is allowed by the standard to not have to consider the case where a function pointer points to a regular variable. So it does not make the connection that the argument to mpu_fault_test() is actually used by said function.

Luckily, fixing this is pretty straightforward. We just have to mark the variable as volatile (and apply the same fix to mpu_fault_test_heap()). That way the compiler isn't allowed to reorder the access in this manner.

Impact of changes

Migration actions required

Documentation

None

---

Pull request type

[X] Patch update (Bug fix / Target update / Docs update / Test update / Refactor)
[] Feature update (New feature / Functionality change / New API)
[] Major update (Breaking change E.g. Return code change / API behaviour change)

---

Test results

[] No Tests required for this change (E.g docs only update)
[] Covered by existing mbed-os tests (Greentea or Unittest)
[X] Tests / results supplied as part of this PR

{{__sync;677f7397-a210-43cf-8efa-d23bf653564a}}
{{__version;1.3.0}}
{{__timeout;20}}
{{__host_test_name;default_auto}}
{{__testcase_count;6}}
>>> Running 6 test cases...
{{__testcase_name;MPU - init}}
{{__testcase_name;MPU - free}}
{{__testcase_name;MPU - data fault}}
{{__testcase_name;MPU - bss fault}}
{{__testcase_name;MPU - stack fault}}
{{__testcase_name;MPU - heap fault}}

>>> Running case #1: 'MPU - init'...
{{__testcase_start;MPU - init}}
{{__testcase_finish;MPU - init;1;0}}
>>> 'MPU - init': 1 passed, 0 failed
<greentea test suite>:0::PASS

>>> Running case #2: 'MPU - free'...
{{__testcase_start;MPU - free}}
{{__testcase_finish;MPU - free;1;0}}
>>> 'MPU - free': 1 passed, 0 failed
<greentea test suite>:0::PASS

>>> Running case #3: 'MPU - data fault'...
{{__testcase_start;MPU - data fault}}
{{__testcase_finish;MPU - data fault;1;0}}
>>> 'MPU - data fault': 1 passed, 0 failed
<greentea test suite>:0::PASS

---