codegen: implement isdefined_tfunc check more similarly to inference

[vtjnash]

No description provided.

[gbaraldi]

I guess this can supercede #52068?

[vtjnash]

Oops, turns out that this optimization is unsound for inference / codegen to attempt since some packages (e.g. NetworkOptions) relies on the non-existence of this optimization--we therefore may need to delete this optimization from inference.

[oscardssmith]

some packages (e.g. NetworkOptions) relies on the non-existence of this optimization

How can they tell whether this optimization was performed?

[vtjnash]

They segfault if the code gets reinferred

[vtjnash]

Part of the cleanup here is perhaps worthwhile (merging code paths into a common helper function), but the main change has been removed from inference for correctness reasons in #53750

[DilumAluthge]

CI is all green here. @vtjnash Is this good to merge now, or are you still waiting on a review from Valentin and/or Gabriel?

[vtjnash]

It is fine now to merge