CUDA: use selected expert cache for IQ2 SSD prefill

[Matthley]

Fixes CUDA SSD streaming for routed MoE models.

1. IQ2/Q2 batch prefill now uses the compact selected-expert table rather than resolving full expert tables.
2. --ssd-streaming-cache-experts is a bounded CUDA LRU, released on model/lifecycle changes.
3. Resident experts are stored in contiguous gate/up/down slabs. When the selected set is resident, kernels receive the slab bases plus remapped stable slot IDs, avoiding per-layer GPU-to-GPU copies into compact staging.
4. CUDA now supports --expert-profile / DS4_EXPERT_HOTLIST to collect routed-expert locality data.

AProjQ4 matched prefill (GB10, e9528de, ctx 1024, forced batch prefill):

• Before: 1.09 t/s; CUDA file cache 78.62 GiB.
• Selected-expert path: 2.59 t/s; CUDA file cache 6.06 GiB.
• Result: 2.38x prefill (+138%) and 92.3% less file cache.

AProjQ8 cache-size tuning smoke (GB10, ctx 1024, deterministic prompt; 22 emitted tokens):

Expert slots	Expert cache	Prefill	Generation	Hits	Misses
256	1.69 GiB	2.85 t/s	1.62 t/s	755	18,056
512	3.38 GiB	2.87 t/s	2.05 t/s	5,027	13,784
1024	6.76 GiB	2.84 t/s	4.02 t/s	11,066	7,745
2048	13.52 GiB	2.65 t/s	6.24 t/s	13,809	5,002

For this model/host, 1024 slots is the measured throughput/memory compromise; 2048 favors generation throughput where the extra memory is available. These are single-run, host- and model-specific tuning observations, not a general performance claim.

Validation:

• make -j4 cuda-spark: PASS
• make test-cuda-q8-scratch: PASS
• Deterministic SSD-streaming smoke: PASS, identical output.