test: setup go-vcr to start recording real providers interactions

[andreynering]

Not a big deal, but in the progress I opened the following PR to go-vcr to allow us to have multi-line formatted strings in the YAML files. This is useful improve legibility, in particular when streaming.

• feat: add WithEncodeOption to tweak yaml marshalling dnaeon/go-vcr#129

[Copilot]

Pull Request Overview

This PR sets up go-vcr for recording real provider interactions in testing, enabling deterministic test replays without making live API calls. The PR introduces a comprehensive test suite for multiple AI providers (OpenAI and Anthropic) with test scenarios covering simple generation, tool usage, streaming, and streaming with tools.

• Added go-vcr dependency and HTTP recording infrastructure
• Created provider tests with support for OpenAI GPT-4o/4o-mini and Anthropic Claude Sonnet models
• Generated test recordings for all supported providers across different test scenarios

Reviewed Changes

Copilot reviewed 18 out of 19 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
go.mod	Adds dependencies for go-vcr and godotenv for test recording and environment management
.gitattributes	Excludes generated YAML test data from diffs and marks as linguist-generated
providertests/.env.sample	Provides template for required API keys
providertests/builders_test.go	Defines language model builders for different AI providers
providertests/provider_test.go	Implements comprehensive test scenarios for AI provider functionality
providertests/recorder_test.go	Sets up VCR recording infrastructure with custom matchers and header filtering
providertests/testdata/**/*.yaml	Generated cassette files containing recorded HTTP interactions for test replay

---

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

[kujtimiihoxha]

Love the way you set it up so it tests the same things for different providers.

This makes sense because we want it to work the same for all language models.

Long term we want to add more tests to make sure that we are parsing the messages correctly and we want to test more multi step scenarios.