FaultManager: Clean up pending_clusters_ when fault is cleared before min_count

[mfaferek93]

Context

Follow-up from #105 (fault correlation engine).

Problem

When a fault is cleared before a pending cluster reaches min_count, the pending_clusters_
map retains a "phantom" reference to the cleared fault code.

Impact

• Low impact - pending clusters expire after window_ms anyway
• Could cause brief inconsistency if cluster becomes active with a cleared fault in its list

Proposed Fix

In CorrelationEngine::process_clear():

1. Iterate through pending_clusters_ (keyed by rule_id)
2. Remove the cleared fault_code from each pending cluster's fault_codes list
3. Reassign representative if the cleared fault was the representative

Acceptance Criteria

• Cleared faults are removed from pending_clusters_
• Representative is reassigned if needed
• Unit test covering this scenario