fix: unfulfilled nested dead code lint

[dronavallipranav]

Fixes #160942

Dead code diagnostic being covered by parent's dead diagnostic resulting in false positives on expectations not being fulfilled. fixed by checking each child of a dead item for its own expectation now and fulfilling their expectation (if valid). Follows suite for manually fulfilling expect level diagnostics (copies mechanism from LintContext::fulfill_expectation.

UI tests fail as expected on main (without fix)

Input

// This struct is only used on windows.
#[cfg_attr(not(target_os = "windows"), expect(unused))]
struct Foo {
    // This field is never used regardless of whether on windows or other.
    #[expect(unused)]
    x: u32,
}

Output
Before

warning: this lint expectation is unfulfilled
 --> src/lib.rs:5:14
  |
5 |     #[expect(unused)]
  |              ^^^^^^

After: no output

I used llm originally to get an understanding of the existing flows and context of the flow around the problem, through prompting helped locate how could reuse the dummy diagnostic lint strategy created cases and I worked backwards from these cases to generate implementation and corresponding logic. I then had it wrote UI tests for those cases for me. The UI tests felt redundant to recreate since I already had it generate the original cases the linter was meant to handle and after review they seem robust couldn't seem to come up with another case.

[rustbot]

r? @folkertdev

rustbot has assigned @folkertdev.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 75 candidates
• Random selection from 17 candidates

[folkertdev]

r? JonathanBrouwer because lints?

I'm not sure we even want this though, it probably needs T-lang approval, or input at least