fix: correct device guards and reject pool growth during capture

[gongchensu]

Summary

• Validate enabled device types before dispatch in include/infini/rt/device_guard.h, and restore the previous device even when the guard initially selects the already-current device.
• Reject CUDA pool growth during native stream capture in src/reusable_memory_pool.cc, while allowing cached-block reuse. Document the preallocation requirement in include/infini/rt/reusable_memory_pool.h.
• Add kEventNotReady mappings for Iluvatar, Hygon, T-Head, MetaX, Mars, and Moore so the existing generator can emit their IsEventNotReady dispatch cases.
• Add DeviceGuard and CUDA pool capture regression tests, and extend native runtime tests with event-status classification checks.

Base: refactor/modern-infini-stack-review@b413816.
Head: fix/modern-infini-stack-review@86b2326.

Motivation

The refactor branch introduces runtime utilities with several boundary cases:

• A backend that was not built can assert in Debug or be incorrectly reported as selected in Release, allowing a pool to report a different device from the actual allocation backend.
• A guard that starts on the requested device skips restoration even if code inside its scope later changes the device type or index.
• A pool cache miss during CUDA capture creates graph-owned allocation nodes without the corresponding replay lifetime management; subsequent graph launches can fail. Rejecting growth before allocation provides an immediate error and preserves capture for cached blocks.
• Event-capable backends without kEventNotReady fall through to false in the generated public helper, misclassifying a normal pending event.

Type of Change

• feat - new feature / new backend capability / new public API
• fix - bug fix
• perf - performance improvement without behavior 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)
• Hygon (WITH_HYGON)
• MetaX (WITH_METAX)
• Moore (WITH_MOORE)
• Cambricon (WITH_CAMBRICON)
• Ascend (WITH_ASCEND)
• Build system / CMake / generated headers
• Public headers / installed consumer API
• Documentation only

Smoke Build and Test Result

paste smoke build and test output here

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
CPU
NVIDIA
Iluvatar
Hygon
MetaX
Moore
Cambricon
Ascend

Full `ctest` output (optional)

paste here

Benchmark / Performance Impact

Notes for Reviewers