Fix ta dualstack

[Horiodino]

Description:
This Pr fixes the Target Allocator generated duplicated targets when DualStack (IPv4 and IPv6) EndpointSlices referencing for the same service.
This update added a logic to detect and skip duplicates based on a service & pod combination key, ensuring that each target is discovered only once per pod while preserving distinct targets across different pods or services.

Link to tracking Issue(s):

• Resolves: Target Allocator Generate duplicated Target/Metrics when Service is Dual Stack #4494

Testing:

• Added unit tests , all existing e2e passes locally.

Documentation:

[github-actions]

E2E Test Results

 34 files  ±0  227 suites  ±0   1h 56m 59s ⏱️ - 4m 31s
 90 tests ±0   90 ✅ ±0  0 💤 ±0  0 ❌ ±0 
231 runs  ±0  231 ✅ ±0  0 💤 ±0  0 ❌ ±0 

Results for commit a5f5396. ± Comparison against base commit 643592c.

♻️ This comment has been updated with latest results.

[Horiodino]

cc @frzifus FYI, you can follow the discussion in the issue. I’ve paused this for now. If the proposed approach looks good, you can have some discussion with swiatekm ,then we will move ahead but the core problem still remains because Prometheus (via Prometheus Operator) defaults to the Endpoints role, while the Target Allocator currently uses EndpointSlice discovery (hard-coded).
This difference means TA will continue to surface dual-stack slices separately unless we explicitly handle it.