fix(qwen3.5-moe): sanitize toward the gate/up/down expert names SparseMoeBlock actually loads

[inureyes]

Summary

Every qwen3.5-MoE / qwen3.6-MoE checkpoint has failed to load since PR #588 with Missing weight: model.layers.0.mlp.switch_mlp.gate_proj. Found by the 2026-07-06 full benchmark sweep: qwen3.5-35b-a3b-4bit and qwen3.6-35b-a3b-4bit benched at 70 / 67 tok/s on 2026-06-12 and were FAIL:bench on 2026-07-06.

Root cause

PR #588 renamed the shared qwen3_next SwitchGLU / SparseMoeBlock expert projections from w1/w2/w3 to gate_proj/up_proj/down_proj, but qwen3_5::sanitize_weights kept normalizing in the OLD direction: step 6 stacked per-expert experts.{e}.{w1,w2,w3} into switch_mlp.{w1,w2,w3}, and step 8 renamed pre-stacked switch_mlp.{gate,up,down}_proj INTO switch_mlp.{w1,w3,w2}. The block then requests switch_mlp.gate_proj and finds nothing. PR #587's per-expert gate/up/down stacking pass also targeted the stale w1/w3/w2 slots; its tests asserted sanitize OUTPUT keys only and never loaded the result, so they passed against the wrong convention — the same "synthetic passes, real checkpoint fails" class as #588 root-caused.

Fix

All three passes now align with the block naming: per-expert w1/w2/w3 stacks into gate_proj/down_proj/up_proj (w1→gate, w2→down, w3→up), per-expert gate/up/down stacks into the same-name slots, and step 8 renames legacy stacked w1/w3/w2 FORWARD into gate/up/down while pre-stacked gate/up/down checkpoints pass through untouched.

Validation

• The missing gate is now a regression test: sanitize output must load end to end through the shared SwitchGLU for all four naming conventions checkpoints ship (per-expert and pre-stacked, in both namings). It fails on the pre-fix sanitize by construction (the old direction produces w1 keys the block never requests).
• The three fix(qwen3.5-moe): stack per-expert gate_proj/up_proj/down_proj weight names #587 tests are retargeted at the gate/up/down expectations and pass.
• Real checkpoints: qwen3.5-35b-a3b-4bit and qwen3.6-35b-a3b-4bit load and generate again (57.8 / 72.8 tok/s; "The capital of France is Paris." at a 200-token budget through the thinking template); dense qwen3.5-4b-4bit unaffected.
• cargo fmt, cargo clippy --all-targets -- -D warnings, and the qwen3_5 unit suite are clean. The two models will be re-added to the 2026-07-06 benchmark CSV after merge.

Closes #670