test(kademlia): deflake TestAddressBookQuickPrune and TestOutofDepthPrune

[martinconic]

Checklist

• I have read the coding guide.
• My change requires a documentation update, and I have done it.
• I have added tests to cover my changes.
• I have filled out the description and linked the related issues.

Description

Deflakes two timing-sensitive tests in pkg/topology/kademlia that intermittently fail in CI on unrelated branches:

• TestAddressBookQuickPrune — e.g. kademlia_test.go: timed out waiting for counter ... got 3 want >= 4
• TestOutofDepthPrune — e.g. kademlia_test.go: bin 5, got 14, want more than 16

Both failures are test-harness timing bugs: the tests gated async connect/prune progress on fixed time.Sleep barriers and on the assumption that each Trigger() yields exactly one dial, which breaks under load (a trigger landing inside a peer's TimeToRetry window is skipped; a 500ms barrier is not always enough on a loaded runner).

Product change required by this approach. synctest demands that every goroutine started inside the bubble terminates before the test ends. Kademlia's metrics collector opens an in-memory leveldb in New (shed.NewDB), but the *shed.DB handle was passed to the collector and then dropped — Close never closed it. This is a real (if low-severity) resource/goroutine leak that was masked in tests by the goleak ignores in main_test.go. We now retain the store handle on the Kad struct and close it in Close, after the collector is finalized. With the store closed, the leveldb compaction/session goroutines exit.

AI Disclosure

• This PR contains code that has been generated by an LLM.
• I have reviewed the AI generated code thoroughly.
• I possess the technical expertise to responsibly review the code generated in this PR.