lldb python formatters: add object validation

[jhelferty-nv]

When providing formatters for types in lldb, we need to iterate through the types to find the appropriate data. If we attempt to iterate into a leaf of an object that isn't valid (e.g., into a nullptr) then python will generate a warning/error message on the command line. If you're using lldb on the command line, and particularly when using the curses-based gui, this results in visible corruption.

This change validates each step along the way to avoid this issue.

(Coded by Claude)

[samestep]

Thanks for working on this! It is really unfortunate that there are so many sharp edges one needs to manually check for when writing LLDB scripts like this, in contrast to the Natvis files in this repo which seem relatively clean.

In general I like the idea of this PR, but I really really don't like this diff as it currently stands :/ it makes the code significantly less readable and maintainable. Is there any way to achieve the same goal in a more maintainable way? Perhaps by putting the validation into some sort of wrapper/helper class/functions, similar to the Children class that already exists here?

In the end though, I'm not actually using these currently and you are, so I'll defer to you on whether to merge this or not.

[jhelferty-nv]

I mean, those are fair criticisms. I could try and push checks into a wrapper class. (It feels like there should be some python-y way to abstract that a bit using, I dunno, @decorators or something.) I think we'll still end up with branching in the actual formatters, but it might be reduced a bit.

Ideally we wouldn't need to hand-code any of this. My preferred solution is to have something that uses the .natvis files to generate the python formatters for us, instead of having to maintain the python formatters directly.