docs: add ChatOpenAI Custom setup guide

[Leeon-Tang]

Summary

Adds a ChatOpenAI Custom documentation page for OpenAI-compatible chat completion providers.

Changes

• Adds ChatOpenAI Custom setup steps under LangChain chat models
• Documents how to configure WaveSpeed LLM through the OpenAI-compatible base path
• Replaces the existing ChatOpenAI Custom broken reference in the chat models list
• Adds the new page to the English summary navigation

Validation

• git diff --check HEAD~1..HEAD
• Local Markdown relative-link check for en/SUMMARY.md and en/integrations/langchain/chat-models/README.md
• Smoke-tested WaveSpeed's OpenAI-compatible chat completions endpoint with https://llm.wavespeed.ai/v1
• Verified anthropic/claude-opus-4.6, deepseek/deepseek-v4-pro, and deepseek/deepseek-v4-flash each returned valid chat completion content across 3 consecutive runs
• Re-tested the documented example deepseek/deepseek-v4-flash after review updates and confirmed it returned finish=stop

Note: I kept deepseek/deepseek-v4-flash as the documented example because it avoids vendor-version ambiguity and returned stable completions in testing.

[gemini-code-assist]

Code Review

This pull request introduces documentation for the ChatOpenAI Custom node, including its addition to the global summary and the chat models README. The feedback suggests correcting a non-existent model version used in the examples and generalizing the troubleshooting steps to be applicable to all providers rather than being specific to WaveSpeed.

[gemini-code-assist]

The model version 4.6 for Claude Opus does not exist. Anthropic's current models are in the 3.x series (e.g., Claude 3 Opus, Claude 3.5 Sonnet). Using a non-existent model version in the documentation can be confusing for users. Please use a real model ID as an example.

Suggested change

	| Model Name | A WaveSpeed model ID such as `anthropic/claude-opus-4.6` |
	| Model Name | A WaveSpeed model ID such as anthropic/claude-3-opus |

[gemini-code-assist]

The troubleshooting steps are currently specific to WaveSpeed. Since this is a general guide for the ChatOpenAI Custom node, it is better to use generic language that applies to any OpenAI-compatible provider, or move these specific points under the WaveSpeed section to avoid confusion for users of other providers.

Suggested change

	* If requests return `401 Unauthorized`, confirm that the credential contains a WaveSpeed API key.
	* If requests return `404 Not Found` or `model not found`, use the full `vendor/model` ID shown in the WaveSpeed catalog.
	* If requests fail to connect, confirm that **Base Path** is exactly `https://llm.wavespeed.ai/v1`.
	* If requests return 401 Unauthorized, confirm that the credential contains a valid API key for your provider.
	* If requests return 404 Not Found or model not found, use the full vendor/model ID as required by your provider.
	* If requests fail to connect, confirm that the Base Path is correctly configured for your provider's endpoint.