pi-synthetic v0.29.2

6a74f95: Sync the static model catalog with the Synthetic API: add hf:deepseek-ai/DeepSeek-V4.1-Flash-1017, a pinned 10/17 snapshot of DeepSeek-V4.1-Flash with vision support, 524K context, and DeepSeek-tier pricing ($0.6/M input, $1.2/M output, $0.03/M cache reads). Runtime-verified: reasoning_effort: "none" disables reasoning (exposed as the off level) and advertised efforts none/low/high/xhigh/max map by identity, identical to the floating model.