feat: support distinct bool tensor dtype in InfiniOps

[baominghelly]

Summary

• Map PyTorch bool tensors to distinct DataType::kBool through the Python and PyTorch C++ bindings, eliminating the bool-as-uint8 workaround.
• Require bool masks in TopkSoftmax and TopkSigmoid; remove the attention-mask dtype override now that native bool mapping exists.
• Add bool coverage for logic/reduction operators and document the dtype.

Motivation

Bool-valued inputs and outputs were previously rejected or represented as uint8 by InfiniOps. This blocked comparison, logical, and reduction tests and made mask contracts inconsistent with PyTorch. Requires the companion InfiniRT bool dtype change (InfiniTensor/InfiniRT#47).

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

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

Built an InfiniOps wheel in the earlier container validation for this same source tree. Retested the final commit against its existing build in baoming_infiniops:

PYTHONPATH=/data/shared/baoming/workplace/new_ops/InfiniOps-bool-dtype-latest/.bool-test-packages python -m pytest -q tests/test_all.py tests/test_any.py tests/test_isin.py tests/test_logical_not.py tests/test_torch_ops.py
3545 passed, 5961 skipped, 106 warnings in 23.94s

The 5,961 skips are existing conditional/parameterized skips, not validated passes. Changed C++ files pass clang-format 21 --dry-run --Werror; git diff --check passes.

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	Yes	Existing wheel and selected CPU/CUDA bool and torch-op tests passed	Full suite not run on final commit
Iluvatar	Yes	Not run - hardware unavailable	Shared dtype/binding path needs validation
MetaX	Yes	Not run - hardware unavailable	Shared dtype/binding path needs validation
Cambricon	Yes	Not run - hardware unavailable	Shared dtype/binding path needs validation
Moore	Yes	Not run - hardware unavailable	Shared dtype/binding path needs validation
Ascend	Yes	Not run - hardware unavailable	Shared dtype/binding path needs validation

Benchmark / Performance Impact

N/A - dtype and mask contract change; no benchmark run.

Notes for Reviewers

• Depends on feat: add distinct bool data type to InfiniRT InfiniRT#47 and an InfiniRT build containing DataType::kBool.
• TopkSoftmax/TopkSigmoid now require a true bool is_padding tensor; previous uint8 callers should be checked. Attention masks no longer need bool-via-byte dtype override.
• Existing branch feat/bool-dtype-latest predates the current branch naming convention (<type>/xxx-yyyy-zzzz); not renamed as part of this PR.
• Other accelerators and full-suite validation remain open; keep this PR in draft pending review.