fix(rpp): Fix device memory leak in HIP histogram_equalize overflow path (#9721)

## Motivation

This PR addresses[ issue 735](https://github.com/ROCm/rpp/issues/735) in
ROCm/rpp: in `hip_exec_histogram_equalize_tensor`, when `required_size`
is greater than SCRATCH_BUFFER_SIZE, the function allocates an overflow
buffer, which does not have an associated `hipFree`. This causes a
device memory leak.

## Technical Details

Wrap the overflow buffer with `std::unique_ptr` with a custom deleter
that calls `hipFreeAsync`.

## Test Plan

CTests should pass.

To verify the overflow issue, and fix, forced the use of the overflow
buffer and then called the function repeatedly (see Details). Without
the fix, there was eventually a `hipMalloc` runtime error on trying to
allocate the buffer with status `hipErrorOutOfMemory`. With the fix,
there was no error.

**Details**
Around line 390 in `histogram_equalize.cpp`, temporarily changed the `if
else` on whether to use the preallocated scratch buffer or allocate the
overflow buffer to always allocate the overflow buffer.
Full command for testing the overflow:
`./Tensor_image_hip
$ROCM_PATH/share/rpp/test/TEST_IMAGES/three_images_mixed_src1
$ROCM_PATH/share/rpp/test/TEST_IMAGES/three_images_mixed_src2
/tmp/test_out 0 0 102 0 100000 1 0 0 0 0 3 0 0 0 0 /tmp`

## Test Result

    Start 6: rpp_qa_tests_tensor_image_hip_all
1/1 Test #6: rpp_qa_tests_tensor_image_hip_all ...   Passed  1313.26 sec

100% tests passed, 0 tests failed out of 1

Total Test time (real) = 1313.26 sec

## Submission Checklist

- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

---------

Co-authored-by: Kiriti Gowda <kiritigowda@gmail.com>