Allow present_key to be empty when past_key is provided in Attention

[justinchuby]

The original check enforces both the present_key and the past_key must be present. But with IO-binding there may be an issue: The past_key can be nullptr even when present_key is allocated. In reality, the kernel should just do the computation when it has the data, or when the output is requested.

[Copilot]

Pull Request Overview

This PR relaxes the validation logic for past/present key tensor requirements in the Attention operator. Previously, the code enforced that past_key and present_key must both be null or both non-null. The new logic allows present_key to be empty when past_key is provided, accommodating ONNX models where dead code elimination may remove unused outputs.

Key changes:

• Replaced strict bidirectional enforcement with unidirectional validation
• Removed implementation limitation comments that are no longer applicable

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[gramalingam]

There must be logic down below that concatenates past key to generate present key which will be impacted: because with this generalization you will still need to construct the concatenated present key which is used in the attention computation, but there will be no present-key-buffer to store this concatenated value; it will implicitly use the nullptr for this buffer or some other error will happen

[gramalingam]

So, the deleted comment above pretty much says the same thing

[justinchuby]

I updated the check to match the (recovered) comment so that it does not prevent the case when past_key is null, seen with io binding.

[justinchuby]

I believe line 493-497 handles it?