Convert getTifBoundary.m to Python

[dlee489]

Translate the logic in getTifBoundary.m to python.

[dlee489]

On local environment, I pass all the test cases, while on the CI environment, I consistently fail the test, as shown in this commit: 4579201. Slightly confused as to what to do about this, I looked at the dependencies and they seem the same, the coding AI's say to loosen the tolerances because of different operating systems.

Any insight?
@Phil-Dua
@yuminguw

On local:

[yuminguw]

All passed on my local Windows WSL environment. But I observed some dependency issues, ie. the requirements defined in the configuration file are not satisfied or incompatible, perhaps we need to update the configuration file.

[dlee489]

Seems like an issue with how numpy's KNN works based on the OS -- since it works on both Yuming's local environment (Linux) and my local environment (MacOS), I'll move on to #6, and Yuming will look over the code to make sure it's working correctly.

[hinerm]

@dlee489 I used the "Explain error" button in the failed job and I agree with its assessment. I am very curious about the native back end like OpenBLAS.

Here were its suggestions for capturing as much info as possible:

Capture environment info at the start of the CI job (add to test setup or job step):

    uname -a; cat /etc/os-release
    lscpu; cat /proc/cpuinfo
    git rev-parse HEAD
    Show language/runtime versions:
        Python: python -V; pip freeze > pip-freeze.txt
        Node: node -v; npm ls --depth=0
        Java: java -version
    Show numeric/lib backend:
        Python + numpy: python - <<'PY' import numpy as np, sys print("numpy", np.version) np.show_config() print("python", sys.version) PY

Capture relevant env vars that affect determinism:

    printenv | grep -E "OMP|OPENBLAS|MKL|VECLIB|PYTHONHASHSEED|TZ|LANG|LC_"
    In tests, print RNG seeds used.

Force single-threaded math in CI to see if results change:

    export OMP_NUM_THREADS=1 OPENBLAS_NUM_THREADS=1 MKL_NUM_THREADS=1 VECLIB_MAXIMUM_THREADS=1
    Re-run tests; if result matches local, BLAS/threading is likely the cause.

[dlee489]

This issue is now involved with #22, in the same branch.