fix(ext/ai): Invalid tensor length causing Segmentation Fault

[kallebysantos]

What kind of change does this PR introduce?

Bug fix, CRITICAL

What is the current behavior?

If a Tensor received from JS land haves data.length not compatible with the dims then when feeding the Session a Segmentation Fault: 11 error will be panic!.

How to reproduce it:

The tensors is successfully handled by rust as well JsTensor::extract_ort_input but error will be panic when feed this values into the Session.

const t1 = new Tensor("float32", [], [1] ); // This one always will result in panic!
const t2 = new Tensor("float32", [0.1, 0.2, 0.3], [1, 2] ); // Session may safely throws while validating inputs
// ...
await core.ops.op_sb_ai_ort_run_session("123", { t1, t2 });

./scripts/run.sh: line 13: 12963 Segmentation fault: 11  RUST_BACKTRACE=full ./target/debug/edge-runtime "$@" start --main-service ./examples/main --event-worker ./examples/event-manager

What is the new behavior?

Following the ort api reference: The 'N' of elements 'T' must be equal to the Tensor's 'dims product'.
So we must prevent it by checking if the Tensor length reflects in its dim shape

[laktek]

@nyannyacha can you take a look?

[github-actions]

🎉 This PR is included in version 1.66.7 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀