fix: read cassette test sources as utf-8

[Ghraven]

What changed

• Read Python test files in scripts/check_cassettes.py with explicit encoding='utf-8' before AST parsing.
• Added a focused test that parses a VCR-marked test file containing non-ASCII source text.

Problem

Path.read_text() uses the platform default encoding when no encoding is provided. On non-UTF-8 locales, the cassette checker can fail before AST parsing if a test file contains UTF-8-only text.

Before/after

Before: the cassette checker depended on the process default encoding when reading test source files.
After: it consistently decodes test source as UTF-8, matching the repository's Python source encoding expectations.

Verification

• Direct helper verification with a temporary UTF-8 test file containing Olá, 世界
• Attempted python -m pytest tests/test_check_cassettes.py, but this local lightweight environment is missing optional dependencies needed by the repo-level pytest config (opentelemetry import while resolving warning filters), so pytest stopped before collecting the new test.