Fix portable fast-path predicates

[voltjia]

Summary

• Render Python-backend contiguous fast-path conjunctions as nested binary boolean operations.
• Preserve the existing CUDA C-style conjunction output.
• Cover Triton, TileLang, and CUDA source structure with one regression test.

Root cause: the shared SSA emitter produced one three-value Python BoolOp; Triton 3.1 and 3.3 accept only binary boolean operations in this path.

pytest output:

NVIDIA (accelerator-dev/nvidia:latest), affected suites:
Running 31 items in this shard
...............................                                          [100%]
31 passed in 6.47s

NVIDIA, exact renamed selector:
Running 1 items in this shard
.                                                                        [100%]
1 passed in 1.12s

Iluvatar (image e8382905ffaa), regression plus add integration:
..                                                                       [100%]
2 passed in 2.69s

Hygon (sglang-deepseek-v4-dev-zkjh), master reproduction:
FAILED tests/test_add.py::test[98432-dtype0-cuda]
UnsupportedLanguageConstruct: chained boolean operators are not supported
1 failed in 14.39s

Hygon, exact head 2d720193, affected suites:
.............................                                            [100%]
29 passed in 16.74s