Add NN format samples builder

[vshcryabets]

No description provided.

[Copilot]

Pull Request Overview

This PR adds neural network format sample builders for training machine learning models on AST-based code formatting. The implementation includes support for generating training data from abstract syntax trees and preparing it for LSTM-based neural network training.

• Adds neural network sample builder infrastructure with dictionary and sequence management
• Implements tree-to-vector conversion for RNN training data generation
• Integrates file metadata management in code generators for enhanced tracking

Reviewed Changes

Copilot reviewed 77 out of 77 changed files in this pull request and generated no comments.

File	Description
keras/format/	Neural network training data, model files, and Python utilities for LSTM-based code formatting
cgen-nnparser/	Core domain logic for NN format projects, dictionaries, and tree processing
cgen-lib/	Enhanced file generators with metadata support and updated test expectations
cgen-console/	Added command-line option for storing formatted output trees

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}