Add OpenAI API Support

[ib-bsb-br]

This pull request introduces two major updates to the messy-folder-reorganizer-ai tool:

OpenAI API Integration: Provides users with the option to use OpenAI's models for language processing (folder naming) and generating embeddings. This serves as an alternative to the existing local Ollama setup, offering greater flexibility and access to different AI capabilities.

Release Workflow Refinement: Streamlines the GitHub Actions release workflow. These changes aim to improve maintainability, simplify the release process by focusing on Linux builds, and ensure correct binary naming for release artifacts.

🚀 OpenAI API Integration
This enhancement allows users to choose their preferred AI provider, either local via Ollama or remote via the OpenAI API.

Key Changes & Features:

New CLI Argument --ai-provider:

Accepts local (default, for Ollama) or openai.

OpenAI Specific CLI Arguments:

--openai-api-key: Your OpenAI API key. This can also be set via the OPENAI_API_KEY environment variable.

--openai-llm-model: Specifies the OpenAI model for folder name generation (default: gpt-4o-mini).

--openai-embedding-model: Specifies the OpenAI model for embeddings (default: text-embedding-ada-002).

--openai-api-base: Optional custom base URL for OpenAI-compatible APIs (default: https://api.openai.com/v1).

--openai-temperature: Optional sampling temperature for OpenAI LLM.

--openai-max-tokens: Optional max completion tokens for OpenAI LLM.

--openai-embedding-dimensions: Optional output dimensions for newer OpenAI embedding models.

Ollama Specific CLI Arguments (Usage Clarified with --ai-provider):

--language-model (or -L): Specifies the Ollama LLM model name. This argument is required when --ai-provider is set to local.

--embedding-model (or -E): Specifies the Ollama embedding model name. This argument is required when --ai-provider is set to local.

--ollama-server-address (or -n): Specifies the Ollama server address (default: http://localhost:11434), used when --ai-provider is local.

Core Logic Updates:

Introduced src/ai/openai_client.rs and src/ai/openai_protocol.rs to manage interactions with the OpenAI API.

Refactored src/ai/embeddings_request.rs (function get_embeddings to get_ai_embeddings) and src/ai/llm_request.rs (function ask_ai_for_reordering_plan to get_ai_reordering_plan) to dynamically call the appropriate AI provider based on user selection.

Updated src/configuration/args.rs to include the new AiProvider enum and all associated arguments.

Enhanced src/errors/app_error.rs with specific error types for OpenAI (e.g., OpenAIAPIError, OpenAIAuthError) and for missing Ollama model configurations when local provider is selected.

Modified src/commands/process.rs to validate required arguments based on the selected AI provider.

Documentation (README.md):

The main description now highlights the dual AI provider capability.

A new "AI Provider Configuration" section has been added, providing detailed instructions and examples for setting up and using both Ollama and OpenAI.

The "Command-Line Arguments" table has been comprehensively updated to reflect all new and modified arguments.

⚙️ Release Workflow Refinement (.github/workflows/release.yml)
This section details refinements to the GitHub Actions workflow used for creating releases. These changes focus on enhancing maintainability, simplifying the process by targeting Linux builds, and correcting binary naming conventions.

Key Changes:

Targeted Linux Build: The workflow now exclusively builds on ubuntu-latest, removing the previous multi-OS build matrix (which included macOS variants). This simplifies the release process and focuses efforts on a primary platform.

Binary Naming Correction: The output binary name in the release archive is now consistently messy_folder_reorganizer_ai (using underscores), aligning with the fix in commit Fix(workflow): Update binary name to use underscore and resolving previous inconsistencies.

Toolchain Update: Switched from dtolnay/rust-toolchain@stable to actions-rs/toolchain@v1 for Rust environment setup.

Simplified Build Command: The cargo build --release command no longer specifies a --target, relying on the native target of the ubuntu-latest runner.

Archive Naming: The generated .tar.gz file is now named messy-folder-reorganizer-ai-${{ env.VERSION }}-x86_64-unknown-linux-gnu.tar.gz, explicitly indicating the Linux target.

Tag Trigger: The workflow is triggered on pushes to tags matching the pattern v* (e.g., v0.1.0).

Affected Files:
.github/workflows/release.yml

README.md

src/ai/mod.rs

src/ai/embeddings_request.rs

src/ai/llm_request.rs

src/ai/openai_client.rs (New)

src/ai/openai_protocol.rs (New)

src/app_core/cluster_processor.rs

src/app_core/destination_processor.rs

src/app_core/sources_processor.rs

src/commands/process.rs

src/configuration/args.rs

src/errors/app_error.rs

[PerminovEugene]

Hey! Thank you for pr, it seems like a good stuff. I am currently on vacation with no laptop. I'll check pr in the beginning of June. Sorry for delay