another try at adding AddAnyAttr to AnyView

[gbj]

@luxalpa had the very reasonable suggestion that perhaps the #[inline(always)] annotation that I've removed here was the problem with the AnyView::add_any_attr() implementation all along. This makes sense to me intuitively as I could see how it would lead to a mutually-recursive inlined function that would be... very big.

I don't have a good test case and can no longer compile the test case that you shared with me @zakstucke, so I'm wondering if some others could try it out to see if this fixes the infinite compile time issues with that change.

See #3156 and discussion beginning here for context.

[zakstucke]

Not looking good :(

After 10 mins of compiling:

cannot encode offset of relocations; object file too large

This error is raised from this compiler code:
https://llvm.org/doxygen/MachObjectWriter_8cpp_source.html

i've also tested by removing all #[inline(always)] across leptos - no bueno either (although it does seem to improve compile times substantially when not adding the AnyView anyattr stuff)

[geoffreygarrett]

Would it be agreeable to feature gate this under any-view-any-attr for now, while we work on getting it fully functional? I haven't encountered any issues on my end.

[zakstucke]

See referenced PR, different implementation but same problem (I was hoping traits and hoisting the downcast out might fix it). It might narrow down the issue.

[geoffreygarrett]

@zakstucke, would you mind giving #3461 a spin on your failing case to see if it helps?

[zakstucke]

No change unfortunately - attempt and testing I did with #3460 seems to show the issue being add_any_attr(attr) rather than nested into_any()

[geoffreygarrett]

Leaving this here for traceability 7fc0c41. Let me what effect it has, if any.