Remove denormalized top-level `uid` field from resourcesv2 collection

[havardelnan]

Context

The resourcesv2 MongoDB collection stores the resource UID in two places:

• metadata.uid — the canonical field from the Go struct (metav1.ObjectMeta.UID)
• uid — a top-level denormalized copy, manually injected in database.go Set() method

These always hold the same value. The top-level uid doesn't exist in any Go struct's BSON tags — it's synthetically added at write time.

Current state

All queries against resourcesv2 filter on the top-level uid, and there's a unique index on it.

Where top-level uid is used:

File	Method	Usage
internal/apiservices/resourcesv2service/database.go:50	Set()	filter: bson.M{"uid": uid}
internal/apiservices/resourcesv2service/database.go:63	Set()	write: doc["uid"] = uid (injection point)
internal/apiservices/resourcesv2service/database.go:88	Patch()	filter
internal/apiservices/resourcesv2service/database.go:267	Del()	filter
internal/apiservices/resourcesv2service/database.go:335	GenerateAggregateQuery()	$in match
pkg/acl/aclstore/scopeexpander_mongo.go:31 (ror lib)	resourceRef struct	projection reads uid
internal/databases/mongodb/seeding/seeding.go:52	Seeding	unique index on uid

Where metadata.uid is used:

File	Method	Usage
internal/apiservices/resourcesv2service/database.go:295	GetHashlistByQuery()	projects $metadata.uid as uid

Proposed changes

1. Change all filters from "uid" → "metadata.uid"
2. Recreate the unique index on "metadata.uid" instead of "uid"
3. Stop injecting doc["uid"] in Set()
4. Update scopeexpander_mongo.go projection in the ror lib
5. Migration: drop old index + remove top-level uid field from existing documents

Benefit

Eliminates data duplication and the fragile manual injection pattern. The canonical Kubernetes metadata UID should be the single source of truth.