Skip quadBroadcast/quadSwap split tests when subgroup_size < 8

[gyagp]

The compute,split tests use the predicate id < subgroupSize / 2, which deactivates the upper half of each subgroup. When subgroupSize is 8 or greater the boundary lands on a 4-lane multiple, so every quad stays fully active and the test exercises legitimate predicated quad operations. When subgroupSize is 4, the only quad gets bisected, leaving no fully active quad — calling quadBroadcast/quadSwap there is undefined behavior and the result is not meaningful to validate.

GPUAdapterInfo.subgroupMinSize is not a sufficient pre-check: the size the implementation actually selects depends on the compiled shader (workgroup shape, register pressure, vendor heuristics), not just the adapter's minimum. Two implementations from the same vendor can report identical {min,max} and pick different runtime sizes for the same shader. Reading @Builtin(subgroup_size) from inside the test shader is the only reliable signal.

Issue: #4650

---

Requirements for PR author:

• All missing test coverage is tracked with "TODO" or .unimplemented().
• New helpers are /** documented */ and new helper files are found in helper_index.txt.
• Test behaves as expected in a WebGPU implementation. (If not passing, explain above.)
• Test have be tested with compatibility mode validation enabled and behave as expected. (If not passing, explain above.)

Requirements for reviewer sign-off:

• Tests are properly located.
• Test descriptions are accurate and complete.
• Tests provide complete coverage (including validation control cases). Missing coverage MUST be covered by TODOs.
• Tests avoid over-parameterization (see case count report).

When landing this PR, be sure to make any necessary issue status updates.

[gyagp]

I changed to use the reported subgroup size instead. Below are the test results on various situations.

@workgroup_size	NVIDIA RTX 5080	WARP 1.0.20	WARP 10.0.26100.8246 (Win11 24H2 built-in)
[4,1,1]	32	4	4
[8,1,1]	32	8	4
[16,1,1]	32	16	4
[32,1,1]	32	16	4
[64,1,1]	32	16	4
[256,1,1]	32	16	4
[4,4,4]	32	16	4

All three report adapter.subgroupMinSize=4 (WARP) or 32 (NVIDIA), but the runtime size for the same shader differs across:
* NVIDIA: always 32 — no skip needed.
* WARP 1.0.20: min(wgThreads, 16) — skip only the subgroup size is 4 (6 of the 46 compute,split cases).
* WARP system Windows 11 24H2: always 4 — every case skips.

[github-actions]

Results for build job (at d75eee4):

 Test case/subcase counts did not change.