CLN: Combine checknull/checknull_old

[lithomas1]

Is there a reason that both checknull and checknull_old are separate functions? It seems like they could be combined into one function with a parameter inf_as_na. This duplication also seems to be present with the isnaobj/isnaobj_old functions.

I'm asking since having this be unified would greatly simplify my work in #42147

[mzeitlin11]

+1 on this definitely, only concern would be if the additional conditional has a perf penalty (since in a bunch of algos checknull is called on every step of loops). Hopefully wouldn't make a difference though since inf_as_na could effectively be known at compile time so the check might be optimized out.

[jbrockmendel]

If perf is OK then id be fine with this.

[mzeitlin11]

Closed by #44541