Fix dataset prefix matching in STAC server

[eloramirez1356]

This pull request improves the accuracy and robustness of dataset resolution for ERA5 and ERA5-Land datasets, ensuring that dataset IDs are matched exactly rather than by prefix. This prevents accidental conflation of similar dataset names (such as precipitation_total and precipitation_total_land). It also adds comprehensive tests to verify this behavior and updates the documentation to clarify how to load these datasets.

Dataset resolution improvements:

• Refactored dataset matching logic in src/stac/stac-server.ts to use a new featureMatchesDataset function, ensuring that datasets are matched exactly rather than by prefix, preventing collisions between similar dataset names (e.g., precipitation_total vs. precipitation_total_land). [1] [2]
• Added a helper function datasetIdFromItemId to extract dataset IDs from item IDs for legacy fallback, ensuring exact matches only.

Testing enhancements:

• Added a new test suite in tests/stac-server.test.ts to verify that only exact dataset matches are resolved and that prefix collisions do not occur. Tests cover both positive and negative cases as well as legacy item ID fallback.
• Improved test setup by cleaning up global stubs after each test to prevent cross-test interference.

Documentation updates:

• Updated README.md to clarify the distinction between ERA5 and ERA5-Land datasets and provided example code for loading each type, helping users avoid dataset confusion.

Test infrastructure:

• Updated test imports to include afterEach and vi for better test isolation and mocking.