[Test] Improve metadata comparison robustness

[coderabbitai]

Context

From PR #34 review discussion: #34

The test uses JSON.stringify() to compare metadata, which can fail if key ordering differs even when the objects are semantically equivalent.

Problem

• Comparison is fragile to key ordering
• Doesn't account for any transformations the resolver might apply to metadata
• Compares against original cell metadata, not resolver output

Impact

Low priority - could cause false positives but hasn't been an issue in practice.

Suggested Enhancement

Use a deep-equal comparison library that ignores key ordering, or normalize/sort keys before comparison.

Backlinks:

• PR [FEAT] Create a Testing Harness #34: [FEAT] Create a Testing Harness #34
• Requested by: @Avni2000