feat: support 2-argument ceil(value, scale) for Spark compatibility

[SAY-5]

Extends ceil to accept an optional scale parameter for Spark compatibility: ceil(3.145, 2) returns 3.15.

Implementation is ceil(value * 10^scale) / 10^scale. Handles Float32/Float64 with Int64 scale, both scalar and array inputs.

Closes #21560

[SAY-5]

The cargo fmt and docs check failures need the full workspace to resolve locally. I can't run cargo fmt with a sparse checkout. If a maintainer could run cargo fmt --all and cargo doc to regenerate the function docs, that should clear both checks. Happy to rebase if needed.

[kumarUjjawal]

The cargo fmt and docs check failures need the full workspace to resolve locally. I can't run cargo fmt with a sparse checkout.

I don't know what that means but you can run the fmt using ci/scripts/rust_fmt.sh similarly there is for the docs update ./dev/update_function_docs.sh.

[SAY-5]

Thanks for the pointer to the scripts. I checked out the full workspace and ran cargo fmt -- no changes needed, the code was already formatted. The docs script needs a full cargo build to generate function tables, which I can't do from a sparse checkout. Could a maintainer run ./dev/update_function_docs.sh to regenerate the math function docs? That should clear the second check.

[SAY-5]

Ran ./dev/update_function_docs.sh locally to regenerate the function docs. The updated scalar_functions.md now includes the 2-argument signature ceil(numeric_expression[, scale]) and example for ceil(3.145, 2) → 3.15. Latest commit regenerates all three doc files (aggregate, scalar, window).

[coderfender]

@SAY-5 , could you run cargo fm --all on your repo and see if that doesnt help ?

[Jefffrey]

This PR seems to be modifying the datafusion version of ceil and not the datafusion-spark function. Looks like we have a PR for the datafusion-spark version here:

• [datafusion-spark] Support 2-argument ceil(value, scale) #21710

[SAY-5]

@Jefffrey thanks — you're right, I was modifying the core datafusion ceil and not the datafusion-spark function. Closing in favour of #21710 which addresses the right module. Apologies for the noise.