PyTorch-compatible backward API

[tohtana]

Currently DeepSpeed's backward API has more constraints compared to PyTorch's normal backward API.
Here is the usage as described in the documentation:

    loss = model_engine(batch)
    model_engine.backward(loss)

In this example,

1. Only accepts a (scalar) loss value
2. Need to call engine's backward API

In contrast, in standard PyTorch, you can do:

    output = model(batch)
    output.backward(out_grad)

There are several use cases that rely on this flexibility. For example, combining multiple models or using loss functions defined separately from the main model.

If you attempt the same pattern with a DeepSpeed engine, some preprocessing and postprocessing steps will be silently skipped, which can lead to incorrect results.

The document explains we can call _backward_epilogue manually (possibly backward_prologue as well). However, it's easy for users to miss these calls, and passing a non-scalar gradient is still not supported.

This PR introduces the same .backward() behavior as PyTorch, allowing .backward() to be called directly on tensors and supporting non-scalar outputs.

To implement post-backward hooks, we had to use some torch internal APIs. See comments for more details. When the internal APIs are not available, DeepSpeed engine only accepts the traditional way model_engine.backward(loss).

[sfc-gh-truwase]

@tohtana, this is a very exciting usability improvement. Please remember to update the documentation.

[tohtana]

@sfc-gh-truwase I think this PR is now ready for review, though the latest change on HF transformer causes an error with test_zero_nesting_init.py::TestNestedParallelInit::test_nested_parallel_init.