refactor(ops): migrate conv2d to canonical InfiniOps API

[voltjia]

Summary

• migrate the InfiniCore Conv2d adapter from deprecated ConvInfinilm to canonical Conv2d
• map the existing convolution arguments to the canonical stride, padding, and dilation order
• keep the InfiniCore public API, graph planning, and execution behavior unchanged

API alignment

InfiniCore adapter	Previous InfiniOps API	Canonical InfiniOps API	Alignment target and evidence
Conv2d	ConvInfinilm::Call(handle, config, input, weight, bias, pads, strides, dilations, groups, out)	Conv2d::Call(handle, config, input, weight, bias, stride, padding, dilation, groups, out)	PyTorch Python API: torch.nn.functional.conv2d(input, weight, bias=None, stride=1, padding=0, dilation=1, groups=1). InfiniOps preserves the Python argument order and adds the explicit trailing output required by its C++ input-attribute-output convention.

Dependency

This PR is based on refactor/migrate-paged-caching-infiniops / #1467 and advances the InfiniOps gitlink from e733e325 to 1ecce783, the head of InfiniOps #887. The canonical convolution schemas landed in InfiniOps #882; #887 supplies the focused metadata-view compatibility fix needed to build them against the current InfiniRT API.

Against #1467, this PR changes one adapter file and the InfiniOps gitlink.

Validation

Validated commit a35c077b remotely on ssh nvidia with accelerator-dev/nvidia:latest, using the exact InfiniOps #887 head 1ecce783 in a temporary integration merge:

• built and installed _infinicore with the canonical conv2d provider in the wrapper allowlist
• ran python3 test/infinicore/ops/conv2d.py --nvidia: 12/12 passed
• clang-format 16.0.6 --dry-run --Werror passed for the changed C++ file
• git diff --check passed

[voltjia]

Superseded by #1480, which consolidates the canonical InfiniOps adapter migrations while preserving each logical change as a separate commit.