fix(hip): share Gather and tensor-shape contracts

[fhanuman]

Summary

Unify result-shape construction, reification, and static verification for Gather-family and tensor-shape operations.

• share Gather and GatherND dimension-splicing rules across converters, reifiers, and verifiers;
• define GatherElements output shape from its indices operand through the named same-shape contract;
• validate Transpose permutations through one shared rule before materializing dimensions;
• define Resize output N/C from the input and static spatial extents from the imported result template;
• define Size as a rank-zero i64 result through a shared semantic rule;
• compute GatherBlockQuantized output shapes from the logical dequantized tensor rather than physical packed-byte storage;
• account for packed int4 by doubling a surviving quantize-axis extent and keep the converter, lowering, runtime interpretation, and tests in the same layer.

GatherND retains its DPS-init fallback when the runtime tuple width prevents static output-rank inference.

Stack

This is PR 9 of 12 and depends on PR 8:

1. Constant carrier/externalization
2. Shape-rule foundation
3. Reshape provenance
4. MatMul/Gemm
5. Broadcast/Reduction
6. Loop carriers
7. Payload/Readback
8. Conv/Pool
9. Gather/Tensor ← this PR
10. Attention/Normalization
11. Explicit contract enforcement
12. Refinement hardening

AI assistance

AI tools assisted with the shape-contract refactoring and stack reconstruction. The changes were checked with the local build, focused Gather/Transpose/Resize/Size LIT, and targeted GPU-vs-CPU tests.

Made with Cursor

[github-actions]

Thanks for opening a PR!

This project follows LLVM's incremental-development and AI-tool-use
guidance. See CONTRIBUTING.md
for the project workflow.

Before requesting review, please check that:

1. The change is focused. Substantial work links the relevant issue
   or design discussion.
2. The PR documents relevant test results and updates affected
   documentation.
3. If AI tools provided substantial assistance, the description
   explains what was assisted and how it was validated, and commit
   trailers identify the tool. The contributor has reviewed and
   understands the result.

Reviewers are assigned through
CODEOWNERS where ownership
is configured.

[github-actions]

L2 Accuracy Results (EP vs CPU)

Model	Combined L2	Total Elems	Skipped NaN/Inf
conv_test_hybrid	EP inference failed	-	-
GroupQueryAttention_seq256	EP inference failed	-	-
MatMulNBits_o_seq128	EP inference failed	-	-
QMoE_seq128	EP inference failed	-	-

Threshold: 0.01 | Run: 3893 - Commit: 9dd4ecb

[github-actions]

MorphiZen EP Performance Results

Model	QPS	Session (s)	1st Infer (ms)	CPU%	Mem (MB)
full_model_seq128	failed	-	-	-	-
GroupQueryAttention_seq128	failed	-	-	-	-
matmul_down_seq128	failed	-	-	-	-

EPContext Export Performance

Model	QPS	Session (s)	1st Infer (ms)	CPU%	Mem (MB)
full_model_seq128	failed	-	-	-	-

OGA Benchmark Results

Model	Warmup	Reps	Prompt Len	Gen Tokens	TTFT (ms)	TPS	Peak Mem (GB)	GPU Mem (GB)
gpt-oss-20b-webgpu-int4-rtn-block-32	1	5	128	128	-	-	-	0.00
Llama-3.1-8B-awq-g128-int4-asym-fp16-onnx-dml	1	5	128	128	-	-	-	0.00

OGA Wheel Smoke (Python benchmark_e2e.py)

Model	TTFT (ms)	TPS
Llama-3.1-8B-awq-g128-int4-asym-fp16-onnx-dml	failed	-

Run: 3893 - Commit: 9dd4ecb