[mem] Flatten memory checkers into composable per-pool invariant checks

[hnyls2002]

Summary

• Replace 3 monolithic checkers (_check_hybrid_memory, _check_mamba_memory, _check_radix_cache_memory) with composable per-pool checks: _check_full_pool, _check_swa_pool, _check_mamba_pool
• All share a single _check_pool_invariant static method: available + evictable + protected + session_held + uncached == total
• check_memory (idle) uses flat if/if instead of if/elif/else — naturally supports SWA + mamba coexistence
• self_check_during_busy delegates to _check_full_pool(ps, uncached=N) instead of reimplementing the invariant
• Extract _get_total_uncached_size helper from busy check boilerplate
• Watchdog dump_info also uses the flat structure

Follows up on #22554 and #22559.

Test plan

• test_scheduler_pause_generation.py — CPU, mock PoolStats path
• test_metrics.py — exercises idle metrics path
• Mamba / SWA model tests (existing CI coverage)

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[hnyls2002]

/rerun-test test_scheduler_pause_generation.py test_mamba_unittest.py test_metrics.py

[github-actions]

✅ ubuntu-latest (1 test): View workflow run

cd test/ && python3 registered/unit/managers/test_scheduler_pause_generation.py

✅ 1-gpu-5090 (2 tests): View workflow run

cd test/ && python3 registered/unit/mem_cache/test_mamba_unittest.py
cd test/ && python3 registered/observability/test_metrics.py

[hnyls2002]

/rerun-test test_disaggregation_basic.py test_pp_llama.py

[github-actions]

❌ test_disaggregation_basic.py: Ambiguous filename test_disaggregation_basic.py — matched 2 files:

• test/registered/amd/disaggregation/test_disaggregation_basic.py
• test/registered/disaggregation/test_disaggregation_basic.py

Please provide the full path, e.g. /rerun-test test/registered/disaggregation/test_disaggregation_basic.py

❌ test_pp_llama.py: No test file found matching test_pp_llama.py under test/registered/.

[hnyls2002]

/rerun-test test/registered/disaggregation/test_disaggregation_basic.py test/registered/distributed/test_pp_single_node.py

[github-actions]

✅ 2-gpu-h100 (1 test): View workflow run

cd test/ && python3 registered/disaggregation/test_disaggregation_basic.py

✅ 4-gpu-h100 (1 test): View workflow run

cd test/ && python3 registered/distributed/test_pp_single_node.py

[hnyls2002]

/rerun-test test/registered/distributed/test_disaggregation_pp.py test/registered/distributed/test_epd_disaggregation.py

[github-actions]

✅ 8-gpu-h20 (1 test): View workflow run

cd test/ && python3 registered/distributed/test_disaggregation_pp.py

✅ 4-gpu-h100 (1 test): View workflow run

cd test/ && python3 registered/distributed/test_epd_disaggregation.py

[hnyls2002]

/rerun-test test_openai_function_calling.py

[github-actions]

✅ 1-gpu-h100 (1 test): View workflow run

cd test/ && python3 registered/openai_server/function_call/test_openai_function_calling.py