[clang] [analyzer] Fix crash in PaddingChecker on dependent field types (PR #232986)

[Balázs Benics via cfe-commits]

steakhal wrote:

In general, the early return should havr caught this right? So if we would have yet another check, that would confuse the reader why. This is the only reason why I'm hesitant with a workaround. In general Id be fine with a workaround if were under pressure, but the next release is months away.

You don't need to fix it. I think a unittest is constructive enough to ask other folks about this.

https://github.com/llvm/llvm-project/pull/232986