Add Alibaba Coding Plan provider and model configs

[PandaSt0rm]

Summary

Add Alibaba Coding Plan as a new provider using the Model Studio global OpenAI-compatible endpoint, implemented per the official Coding Plan specification:
https://www.alibabacloud.com/help/en/model-studio/coding-plan

A separate provider is required because Coding Plan uses dedicated credentials and base URL, which differ from the existing alibaba provider:

• Existing alibaba provider: https://dashscope-intl.aliyuncs.com/compatible-mode/v1
• Coding Plan: https://coding-intl.dashscope.aliyuncs.com/v1

Changes

• Added providers/alibaba-coding-plan/provider.toml
• Added providers/alibaba-coding-plan/logo.svg
• Added model configs:
  • qwen3.5-plus
  • qwen3-max-2026-01-23
  • qwen3-coder-next
  • qwen3-coder-plus
  • glm-5
  • glm-4.7
  • kimi-k2.5

Validation

• bun validate passes

[PandaSt0rm]

Update in this PR:

• Added MiniMax-M2.5 to Alibaba Coding Plan models.
• Reconciled third-party model limits against Alibaba Model Studio docs.

Limits now set to:

• GLM-4.7: 202,752 context / 16,384 output
• GLM-5: 202,752 context / 16,384 output
• Kimi K2.5: 262,144 context / 32,768 output
• MiniMax-M2.5: 1,048,576 context / 32,768 output

Validation: bun validate passes.

[PandaSt0rm]

Follow-up fix: corrected MiniMax-M2.5 modalities in Alibaba Coding Plan to text-only input.

• before: input = ["text", "image", "video"]
• now: input = ["text"]
• output remains: output = ["text"]

Validation: bun validate passes.

[jnd0]

Adding one more verification datapoint from local testing:

• Ran bun validate on this branch (pass)
• Tested opencode against a generated snapshot from this PR via OPENCODE_MODELS_PATH
• alibaba-coding-plan/* models are discovered and routed correctly

This + #1030 should cover the provider-side fix for anomalyco/opencode#14819 without opencode core hardcoding.