Deprecate vectorized ceil methods in favor of compact broadcast syntax

[Sacha0]

This PR deprecates all remaining vectorized ceil methods (less those for SparseVectors, separate PR) in favor of compact broadcast syntax. Ref. #16285, #17302, #18495, #18512, #18513, #18558, #18564, #18566, #18571 #18575, and #18576. Best!

(Unlike with float, real, etc., the remaining vectorized ceil methods never alias their input. This PR should be less controversial than #18495, #18512, and #18513 as a result.)

[Sacha0]

Travis x64 linux failure seemingly unrelated?

[Sacha0]

Rebased. Best!

[tkelman]

do we now have these structured xdiagonal types returning sparse for general broadcast functions that aren't special cased?

[Sacha0]

do we now have these structured xdiagonal types returning sparse for general broadcast functions that aren't special cased?

Not yet, no. I have not been prioritizing that change given discussion in #18590. Thoughts? Thanks!

[tkelman]

it's what the general sparse types do, right? how hard would it be to plug a few of the structured types into the same mechanism? otherwise will they return dense if you put them through a fused zero-preserving operation? there's a workaround of explicitly converting to sparse, but if there are operations where that's always a better thing to do then return dense, seems we might want to make it automatic

[Sacha0]

Achieving that behavior (if inefficiently, by promoting structured arguments to sparse equivalents prior to broadcast) shouldn't be too difficult. Might take some time though, and seems lower priority than e.g. finishing this series of PRs? Best!

[tkelman]

agreed, material for a separate PR

[Sacha0]

Subsumed by #19791.