[C++][Tensor] Incorrect logic for creating arrow::SparseCOOTensor from Tensor via arrow::SparseCOOTensor::Make (illegal memory access and incorrect values)

### Describe the bug, including details regarding any error messages, version, and platform.

The following method causes illegal memory access when creating from a combination of negative zero and non-zero values, and produces incorrect tensor values (potentially leading to illegal memory access) when creating from a column-major tensor.

https://github.com/apache/arrow/blob/fddd35607ed09ac5c6f1b358c8feb4f207f390f3/cpp/src/arrow/sparse_tensor.h#L583-L594

1- The following code leads to illegal memory access.
```c++
TEST(MyTest, SegFault) {
// clang-format off
std::vector data{
-0.0, -0.0, -0.0,
-0.0, -0.0, -0.0,
-0.0, -0.0, -0.0,
-1.0, -0.0, -0.0,
};

// clang-format on

std::vector shape = {4, 3};
auto buffer = Buffer::FromVector(data);
ASSERT_OK_AND_ASSIGN(auto dense_tensor, Tensor::Make(float32(), buffer, shape));
ASSERT_OK_AND_ASSIGN(auto sparse_coo_tensor,
SparseCOOTensor::Make(*dense_tensor, int64()));
ARROW_LOGGER_INFO("", sparse_coo_tensor->sparse_index()->non_zero_length());
}
```
and the error is:

```
: 1
mimalloc: error: buffer overflow in heap block 0x0200000100C0 of size 152: write after 152 bytes
Process finished with exit code 134 (interrupted by signal 6:SIGABRT)
```
The reason for this is rooted in the following code.

https://github.com/apache/arrow/blob/fddd35607ed09ac5c6f1b358c8feb4f207f390f3/cpp/src/arrow/tensor/coo_converter.cc#L179

In the above code, both positive and negative zero are considered zero; however, in the following code, negative zero is treated as a non-zero value.

https://github.com/apache/arrow/blob/fddd35607ed09ac5c6f1b358c8feb4f207f390f3/cpp/src/arrow/tensor/coo_converter.cc#L193
https://github.com/apache/arrow/blob/fddd35607ed09ac5c6f1b358c8feb4f207f390f3/cpp/src/arrow/tensor/coo_converter.cc#L67
Note that in the above code, the type of` c_value_type` is always an unsigned integer due to the following code.
https://github.com/apache/arrow/blob/fddd35607ed09ac5c6f1b358c8feb4f207f390f3/cpp/src/arrow/tensor/converter_internal.h#L22-L88

2- Incorrect tensor values when creating a SparseCOOTensor from a ColumnMajorTensor via SparseCOOTensor::Make
```c++
TEST(My, ColumnMajor) {
// clang-format off
std::vector data{
1, 4, 7, 10,
2, 5, 8, 11,
3, 6, 9, 12
};
// clang-format on
std::vector data_2{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12};
auto buffer = Buffer::FromVector(data);
auto buffer_2 = Buffer::FromVector(data_2);
std::vector shape = {4, 3};
std::vector strides = {sizeof(int), 4 * sizeof(int)};
ASSERT_OK_AND_ASSIGN(auto tensor, Tensor::Make(int32(), buffer, shape, strides));
ASSERT_OK_AND_ASSIGN(auto tensor_2, Tensor::Make(int32(), buffer_2, shape));
ASSERT_TRUE(tensor->Equals(*tensor_2));
ASSERT_TRUE(tensor->is_contiguous());
ASSERT_TRUE(tensor->is_column_major());
ASSERT_OK_AND_ASSIGN(auto sparse_tensor, SparseCOOTensor::Make(*tensor))
ASSERT_OK_AND_ASSIGN(auto new_tensor, sparse_tensor->ToTensor());
ASSERT_EQ(12, sparse_tensor->non_zero_length());
ASSERT_TRUE(new_tensor->is_contiguous());
ASSERT_TRUE(new_tensor->is_row_major());
// new_tensor is not equal to tensor!!
ASSERT_FALSE(new_tensor->Equals(*tensor));
}
```
`ASSERT_FALSE(new_tensor->Equals(*tensor));` passes because of the following code.
https://github.com/apache/arrow/blob/fddd35607ed09ac5c6f1b358c8feb4f207f390f3/cpp/src/arrow/tensor/coo_converter.cc#L86-L91
Given that logic, the index `{2,3}` is produced for a column-major contiguous tensor with shape `{4,3}`, which is not a correct index.

### Component(s)

C++