ENH: Add integration pytests for `hasi` package

[tbirdso]

• Adds integration pytests for shape analysis pipeline. Each phase (atlas generation, sample correspondence, feature analysis) is organized with its own test depending on output from the previous test.
• Adds pytest-dependency package to manage testing dependencies. Failure at any stage in the integration pipeline will skip subsequent integration tests as we assume input will not be valid.
• Adds existing seaborn and matplotlib dependencies to pyproject.toml

Outlying issues:

• Integration tests can only run on a very small test set due to CI memory constraints, with CI failing due to memory allocation errors when three or more mouse femur images are loaded. Two images is sufficient to do a basic check for shape analysis pipeline functionality. If we want to use a larger test set than this I can generate downsampled test input images to meet the 7GB RAM limit.

[tbirdso]

Working on downsampling images so that we can use a larger data set without exceeding CI memory limits, will update in a subsequent PR.