fix: Do not mutate external TensorInfo in import_memory()

[morgolock]

When a TensorInfo is shared with a Tensor via allocator()->soft_init(), the allocator stores a reference to the caller-provided TensorInfo. If import_memory() is later called, the allocator currently sets TensorInfo::is_resizable(false). Because the TensorInfo is shared, this mutates the caller-owned metadata. Subsequent operations that expect the TensorInfo to remain resizable (e.g. extend_padding()) can fail.

This is observable in multi-threaded or reused inference paths (for example multi-threaded CpuGemmConv2d), where the same TensorInfo is reused across runs. In such cases import_memory() unexpectedly changes the caller-visible TensorInfo state. Fix this by distinguishing between allocator-owned and external TensorInfo instances:

• Track ownership with _owns_info.
• Track imported buffers with _is_imported.
• soft_init() marks the TensorInfo as externally owned.
• import_memory() only mutates TensorInfo::is_resizable() when the allocator owns the TensorInfo.
• When referencing external TensorInfo, allocator logic relies on the internal _is_imported state instead of mutating the caller metadata.

This preserves existing behaviour for allocator-owned TensorInfo while avoiding unexpected mutations of caller-provided metadata.

The change is minimal and does not modify public APIs.

Test:

• Add/enable UNIT test ImportMemoryDoesNotMutateExternalInfo.
• Verify that shared_info.is_resizable() remains true after import_memory() and that extend_padding() succeeds.

Change-Id: Iac9d5807789a8ed74205cdee098b36d0f5269d59

[gunes-arm]

Sorry, these were comments from last week. I forgot to post. Feel free to take it with another reviewer. If not, I'll prioritize this when I get back.

[gunes-arm]

I think this field is extra. From other places, I can see that _info_external is checked against nullptr to decide if the info is owned or not. It seems like having a member function named owns_info that checks _info_external != nullptr is the best way if we need a convenient method.

For example, info is returned based on this check, not the member _owns_info.

TensorInfo &ITensorAllocator::info()
{
    return (_info_external != nullptr) ? *_info_external : _info_owned;
}

[morgolock]

I'm confused, do you want us to remove _owns_info ?

[gunes-arm]

What I'm challenging is why do we need it, given that we have another logic already?

[morgolock]

Good point, just for clarity I suppose. Happy to remove it and just use _info_external != nullptr`

[gunes-arm]

I have some doubts regarding allowing user to extend padding after they import memory. We were disallowing it, but after this change, we're allowing it.

For example, shared_info has been used as a soft initializer for view_tensor. The view_tensor's buffer has been imported from out_tensor, which has a certain size, 16 x 4 x 4 x sizeof(data_type), e.g. 512 bytes if it's Fp16.

Once shared_info's padding is extended, the view_tensor's strides change. If you were to use view_tensor for any purpose after this line, such as printing, filling in with values, feeding it to a function, it won't be any good. It'll potentially cause segfault, or incorrect results being calculated at the least because it'll try to write out of bounds because of the paddings.