feat: add qwen3:8b for DESKTOP-0O8A1RL, update Ollama routing

Benchmarked 2026-05-16 on DESKTOP-0O8A1RL (RTX 5070 Ti Laptop, 12 GB VRAM):
- qwen3:8b: 100% VRAM fit (10.9/10.9 GB) -> 74-86 tok/s
- qwen3:14b: 73% VRAM (11.3/15.6 GB split) -> 17-18 tok/s (4.8x slower)
- qwen3.6:  41% VRAM (11.3/27.5 GB split) -> 17-19 tok/s

qwen3:14b overflows 12 GB VRAM at runtime (9.3 GB GGUF = 15.6 GB loaded).
qwen3:8b fits entirely in VRAM and matches the reference machine speed.

Updated OLLAMA.md: added qwen3:8b to models table, per-machine routing
table, benchmark results. Updated CLAUDE.md model one-liner.
Routing: qwen3:8b for prose on DESKTOP-0O8A1RL, qwen3:14b everywhere else,
qwen3.6 for strict-format tasks on all machines.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

.claude/CLAUDE.md

@@ -272,7 +272,7 @@ Tier 0 — **Ollama is the documentation and classification engine.** Route pros
 | DESKTOP-0O8A1RL | `http://localhost:11434` |
 | Other | `http://100.92.127.64:11434` (Tailscale) |
 
-Models: `qwen3.6:latest` (strict-format: JSON, classification, structured rules, redaction, word-limited summaries, untrusted-input handling), `qwen3:14b` (bulk prose: session logs, commit bodies, free-text drafts — 2x faster), `codestral:22b` (code suggestions — always review). Full reference + routing rationale: `.claude/OLLAMA.md`
+Models: `qwen3.6:latest` (strict-format: JSON, classification, structured rules, redaction, word-limited summaries), `qwen3:8b` (prose on DESKTOP-0O8A1RL — 86 tok/s, full 12 GB VRAM fit), `qwen3:14b` (prose everywhere else — ~66 tok/s), `codestral:22b` (code suggestions — always review). Full reference + per-machine routing: `.claude/OLLAMA.md`
 
 ### GrepAI (Semantic Code Search)