Add Langdock as a supported AI/LLM provider

## Summary
Add native support for Langdock as an LLM provider in plugin-ai (on par with existing providers such as Ollama, OpenAI, Anthropic, etc.), so users can select a Langdock-hosted model for chat completion and AI Agent tasks directly from Kestra flows.

## Use Case
Customer wants to standardize their AI-orchestrated flows on Langdock as the LLM provider, using it the same way they use the existing Ollama provider today: select a model and use it for AI Agent tasks or plain chat completion.

## Current Workaround
Langdock exposes an OpenAI-compatible Chat Completions endpoint (`openai/{region}/v1/chat/completions`, docs: https://docs.langdock.com/en/developer/completion-api/openai). The existing `io.kestra.plugin.ai.provider.OpenAICompliantProvider` can likely be pointed at this endpoint with a Langdock API key for basic chat completion today. This has not been validated for AI Agent / tool-use support, and does not give the first-class model-listing/selection UX that a native provider (like Ollama) has.

## Customer Context
> Pylon ticket: #2102 | Priority: P2

Start by reading the existing io.kestra.plugin.ai.provider.OpenAICompliantProvider and the implementations for providers such as Ollama, then review Langdock's OpenAI-compatible completion API. Done means Langdock is a first-class selectable provider for chat completion and AI Agent tasks, including model listing and tool-use support where applicable.

Written by the indexing model from the issue text.