Add glm-5.3 to alibaba-cn provider

[Spin-Particle]

Adds Zhipu's GLM-5.3 to the Alibaba Bailian pay-as-you-go provider (alibaba-cn, Beijing), using the existing lab entry models/zhipuai/glm-5.3.toml via base_model.

Sources (accessed 2026-09-16):

• https://bailian.console.aliyun.com/cn-beijing?tab=model#/model-market/detail/glm-5.3?serviceSite=asia-pacific-china — Beijing pricing (CNY 8 / 28 / 2 cache hit per 1M tokens, identical to glm-5.2 on the same host, so the same USD conversion 1.1 / 3.851 / 0.275 as the existing glm-5.2.toml entry is used)
• https://help.aliyun.com/zh/model-studio/billing-for-model-studio — per-region price table
• https://docs.bigmodel.cn/cn/guide/models/text/glm-5.3 — model card and capability table

Host-specific deltas vs the lab entry:

• Reasoning: always-on thinking with graded effort. reasoning_effort accepts low, high, max (default max). Thinking cannot be turned off (enable_thinking only accepts true; passing false errors) and GLM-5.3 does not support thinking_budget — so a single effort option, no toggle, no budget. reasoning_content is streamed in deltas (interleaved).
• Cost: same CNY list price as glm-5.2 on Bailian; converted to USD following the existing providers/alibaba-cn/models/glm-5.2.toml convention.
• Limits and modalities are identical to the lab entry (1M context, 131_072 output, text-only), so they are inherited and not restated (override-only).

Not added to alibaba (international) — I could only verify the cn-beijing catalog and pricing; the international gateway can be added separately once its pricing is confirmed.

Context: opencode (and downstream consumers of this API) currently ships a temporary local overlay for this model; once this merges, that overlay retires automatically. Companion PR for kimi-k3 on the same host: #7149 (verified independently — see my comment there).

bun validate passes.

[github-actions]

No actionable findings.