fix(build): separate linked operator allowlist

[voltjia]

Summary

• Adds INFINI_OPS_LINKED_OPS so linked provider selection can be narrower than the generated operator surface.
• Applies the same linked subset to library resolution and wrapper generation while preserving native and ATen implementations for the same operator.
• Documents the configuration and adds regression coverage for default compatibility and mixed native/linked implementations.

Motivation

WITH_LINKED=ON currently reuses INFINI_OPS_OPS for every linked provider. A build that needs native topk_softmax and linked FlashAttention therefore tries to resolve the optional vLLM provider for topk_softmax, even though the native NVIDIA implementation is selected. The separate allowlist keeps the public operator surface intact while resolving only providers the build actually uses.

Type of Change

• feat - new feature / new operator / new platform
• fix - bug fix
• perf - performance improvement (no behavioral change)
• refactor - code restructuring without behavior change
• test - adding or fixing tests only
• docs - documentation only
• build / ci - build system or CI configuration
• chore - tooling, formatting, or other non-code changes
• Breaking change (requires a ! in the Conventional Commits prefix or a BREAKING CHANGE: footer)

Platforms Affected

• CPU (WITH_CPU)
• NVIDIA (WITH_NVIDIA)
• Iluvatar (WITH_ILUVATAR)
• MetaX (WITH_METAX)
• Cambricon (WITH_CAMBRICON)
• Moore (WITH_MOORE)
• Ascend (WITH_ASCEND)
• PyTorch C++ bindings (WITH_TORCH)
• Build system / CMake / CI
• Python bindings / user-facing API

Smoke Test Result

python3 -m pytest tests/test_generate_wrappers.py tests/test_resolve_linked_ops.py -q
51 passed in 1.28s

cmake ... -DINFINI_OPS_OPS=<23 operators> -DINFINI_OPS_LINKED_OPS=flash_attn_varlen_func,flash_attn_with_kvcache
-- Wrapper op allowlist: <23 operators>
-- Linked op allowlist: flash_attn_varlen_func;flash_attn_with_kvcache
-- Resolved 2 linked operator source(s).
-- Configuring done

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	Yes	Configure passed	51 focused wrapper/resolver tests passed; downstream full build pending
Iluvatar	No	N/A - not affected	N/A
MetaX	No	N/A - not affected	N/A
Cambricon	No	N/A - not affected	N/A
Moore	No	N/A - not affected	N/A
Ascend	No	N/A - not affected	N/A

Benchmark / Performance Impact

N/A - build-time provider selection only.

Notes for Reviewers

This PR is stacked on #922 because the downstream integration requires both changes. The commit is independent and can be rebased onto master after #922 merges. With no explicit INFINI_OPS_LINKED_OPS, existing builds keep using INFINI_OPS_OPS for linked providers.