feat(frontend): add support for np.min and np.max

[umut-sahin]

(closes https://github.com/zama-ai/concrete-internal/issues/805)

[umut-sahin]

Btw, I've moved test_min_max.py to test_minimum_maximum.py as they were testing np.minimum and np.maximum, not np.min and np.max. I then introduced test_min_max.py back, so the diff could be a bit confusing. Just ignore the diff and inspect each file separately, it'll be easier.

[BourgerieQuentin]

To add to https://docs.zama.ai/concrete/get-started/compatibility

[bcm-at-zama]

Do we also need to update docs/core-features/minmax.md?

[umut-sahin]

Do we also need to update docs/core-features/minmax.md?

No. That document explains the primitive comparison between two values, like how it's being done. lt's not for the description of high level features.

For that, we're mimicking numpy and it'll be in the compatibility like @BourgerieQuentin requested.

[bcm-at-zama]

Do we also need to update docs/core-features/minmax.md?

No. That document explains the primitive comparison between two values, like how it's being done. lt's not for the description of high level features.

For that, we're mimicking numpy and it'll be in the compatibility like @BourgerieQuentin requested.

but if now, minmax is "native" in CP, do we need to keep a section to explain how it works? ie, if it's all made for the user, we might not explain, there are a lot of things which are done for the user

[umut-sahin]

but if now, minmax is "native" in CP, do we need to keep a section to explain how it works? ie, if it's all made for the user, we might not explain, there are a lot of things which are done for the user

We don't have explicit documentation about np.matmul or np.reshape, why would we have one for min/max? The operation is supported like in numpy, it's enough to document the "primitive" comparison (e.g., strategies for comparing 2 numbers).

[umut-sahin]

I've added extensive documentation of how it works in the code via comments. Could you check @aPere3?

[BourgerieQuentin]

@umut-sahin Check Concrete Python Tests there are a lot of RuntimeError: Transforming FHE boolean ops failed in test_min_max, I don't known why this happening. Must be fixed before merge as well.

[aPere3]

Looks good thanks for the explanations, those are top-notch !

[BourgerieQuentin]

@slab-ci concrete-python-tests-linux