fix min max on zero-sized ndarray

[arcadiaphy]

Description

Calling min or max operator on zero-sized ndarray will return invalid value:

a = mx.nd.zeros((5, 0))
a.min()  # or a.max()
# result
# [-3.4028235e+38]
# <NDArray 1 @cpu(0)>

In numpy, the same situation will trigger ValueError, this PR replicates this behavior by throwing MXNetError.

Edit:
This error will only be triggered in numpy mode:

mx.set_np_compat(True)
a = mx.nd.zeros((5, 0))
a.min()  # or a.max()

Checklist

Essentials

Please feel free to remove inapplicable items for your PR.

• The PR title starts with [MXNET-$JIRA_ID], where $JIRA_ID refers to the relevant JIRA issue created (except PRs with tiny changes)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage:
• Unit tests are added for small changes to verify correctness (e.g. adding a new operator)
• Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore)
• Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL)
• Code is well-documented:
• For user-facing API changes, API doc string has been updated.
• For new C++ functions in header files, their functionalities and arguments are documented.
• For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable
• Check the API doc at https://mxnet-ci-doc.s3-accelerate.dualstack.amazonaws.com/PR-$PR_ID/$BUILD_ID/index.html
• To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change

Changes

• Feature1, tests, (and when applicable, API doc)
• Feature2, tests, (and when applicable, API doc)

Comments

• If this change is a backward incompatible change, why must this change be made.
• Interesting edge cases to note here

[reminisce]

Please note that a zero-size shape represents a unknown shape in mxnet. Please turn on numpy compatibility switch to use 0 as zero dim size.

[reminisce]

It should have already failed here in the infer-shape function of zeros. If you want to use 0 as zero dim size, please turn numpy compatibility by adding a decorator to the test function so that it will fail in ReduceMinMaxAxesShape when a.min() is called.

@with_seed()
@use_np_compat
def test_zero_size_min_max():

[arcadiaphy]

@reminisce You'are right. After #14661, the error I reported will only be triggered in numpy mode.

[reminisce]

test_zero_sized_min_max -> test_zero_size_min_max

[arcadiaphy]

An unrelated custom operator error happens in CI, I'll create another PR to address it.