fix: restore CI for section-based gRASPA parsing - #24

PR #23 switched gRASPA output parsing to section and component headers, but the subprocess fixture still emitted headerless averages. This made the adsorption runtime and pressure-sweep tests fail. The parser change also failed Ruff lint and formatting checks.

This change supplies a shared, structured synthetic log; checks every uptake unit with both fugacity settings; and adds 19 parser regression cases covering framework/excess exclusion, overflow, layout variation, and incomplete output. Public parsing behavior and the compatibility eos argument are preserved.

Validation:

• Focused gRASPA, runtime, and calculation-example tests: 75 passed.
• All existing CI lint/format commands and checks on the updated gRASPA tests passed.
• Source distribution and wheel built successfully.
• GitHub Actions passed all six jobs: lint, build, installed-wheel tests on Python 3.10–3.12, and MCP. CI run