Add GLM-5.2 to wafer.ai provider

[flamerged]

What

Adds GLM-5.2 to the wafer.ai provider catalog. Wafer serves it serverless but it was missing from models.dev, so the opencode CLI (which pulls its provider/model list from https://models.dev/api.json) did not list wafer.ai/GLM-5.2 — only wafer.ai/GLM-5.1.

Verification

Confirmed live on wafer via the unauthenticated models endpoint:

$ curl -sS https://pass.wafer.ai/v1/models | jq '.data[] | select(.id=="GLM-5.2")'
{
  "id": "GLM-5.2",
  "max_model_len": 1048576,
  "zdr_supported": true,
  "wafer": {
    "display_name": "GLM-5.2",
    "tier": "serverless_only",
    "context_length": 1048576,
    "capabilities": { "vision": false, "tools": true, "reasoning": true, ... },
    "pricing": {
      "currency": "usd",
      "input_cents_per_million": 120,
      "output_cents_per_million": 410,
      "cache_read_cents_per_million": 20
    }
  }
}

Pricing & limits (from live wafer /v1/models)

• context: 1,048,576
• output: 131,072
• input: $1.20 / output: $4.10 / cache_read: $0.20 per 1M tokens
• reasoning: true (toggle), tool_call: true, structured_output: true
• attachment/vision: false, text-only I/O

Conventions

Matches the existing providers/wafer.ai/models/GLM-5.1.toml convention: self-contained TOML (no base_model), reasoning_options = [{ type = "toggle" }], underscore-separated numeric literals. Same release_date as the zhipuai base model (models/zhipuai/glm-5.2.toml); last_updated set to today (wafer listed it on 2026-06-21 per the created timestamp in /v1/models).