fix(predicate): track resource UID to handle recreated resources

[doxxx93]

Description

Fixes #1830

Resources recreated with the same name/namespace but different UIDs were incorrectly filtered out by the predicate filter. This PR fixes the issue by tracking UIDs in the cache to distinguish between different resource instances.

Problem

When using predicate filters (e.g., predicates::generation), if a resource is:

1. Created (generation=1, uid=A)
2. Deleted
3. Recreated (generation=1, uid=B)

The recreated resource would be incorrectly filtered out because the cache only used ObjectRef as the key, which deliberately excludes UID from its Hash implementation.

This caused controllers to miss reconciliation of recreated resources, breaking scenarios where resources need to be managed via owner references and garbage collection.

Solution

Created a new PredicateCacheKey struct that includes UID in the cache key:

#[derive(Debug, Clone, PartialEq, Eq, Hash)]
struct PredicateCacheKey {
    name: String,
    namespace: Option<String>,
    uid: Option<String>,
}

Modified PredicateFilter cache from:

HashMap<ObjectRef<K>, u64>

to:

HashMap<PredicateCacheKey, u64>

This ensures different resource instances (with different UIDs) are treated as separate cache entries.

Changes

• Added PredicateCacheKey struct that includes UID in equality/hashing
• Modified PredicateFilter cache to use PredicateCacheKey instead of ObjectRef
• Updated poll_next to create cache keys that include UID
• Added regression test predicate_filtering_should_handle_recreated_resources_with_same_generation

Testing

All existing tests pass (67 passed), including new regression test that verifies:

• First resource with UID "uid-1" passes through
• Recreated resource with same generation but different UID "uid-2" also passes through (previously would be filtered out)

cargo test --package kube-runtime --lib

Checklist

• Fixed the bug described in Recreated objects do not pass through predicate filters #1830
• Added regression test
• All tests pass
• Commit signed with DCO (--signoff)
• No breaking API changes

[doxxx93]

i will fix lint soon

[clux]

thanks a lot for this!

i have one comment initially because it solves it in a way i did not expect, but it's all private implementation details. i just want to make it as simple as possible 😄

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.6%. Comparing base (98bfbe3) to head (8b1b7c7).
⚠️ Report is 7 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #1836     +/-   ##
=======================================
+ Coverage   74.5%   74.6%   +0.1%     
=======================================
  Files         84      84             
  Lines       7856    7877     +21     
=======================================
+ Hits        5849    5871     +22     
+ Misses      2007    2006      -1     

Files with missing lines	Coverage Δ
kube-runtime/src/utils/predicate.rs	77.8% <100.0%> (+7.3%)	⬆️

... and 5 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[doxxx93]

hum.. sorry for lint

[clux]

I think this looks great. I would like one extra thing in the tests ideally, and have a comment suggestion, but tiny things. Thanks a lot. Looks very clean.