Fix entity reload: clear cache before re-applying replay events

[PZahnen]

Problem

Triggering xtractl entity reload left stale entity data in the cache. Two visible symptoms:

1. External entries configured via apiCatalogAdditionalEntries appeared
   multiple times on the API catalog landing page.
2. For some entities the reload also produced a warning followed by a
   Multiple entries with same key: ... exception, e.g.:
   Ignoring 'aaa-suite/alkis-vereinf/services/alkis-vereinf.yml' from FS[.] because it already exists. An entity can only exist in a single source, use overrides to update it from another source.

A web API restart cleared both.

Root Cause

The reload flow emits a synthetic delete event (payload = YAML_NULL) before
the replay event. Two bugs prevented the cache from being cleared:

1. processEvent discarded events with "empty" payloads — but YAML_NULL
   counts as empty, so delete events were silently dropped and
   cache.remove(...) never ran.
2. The follow-up replay event then hit the "single source" warning in
   EntityDataStoreImpl.processEvent, which was missing a
   return List.of();. Despite the "Ignoring..." log message, the event was
   merged into the still-cached value via Jackson's readerForUpdating,
   which appends to lists/maps — causing both the duplicated catalog entries
   and the duplicate-key exception.

Fix

• Let delete events bypass the empty-payload filter so they reach the cache.
• Add the missing early return after the "single source" warning.

Verification

Reproduced locally via the ldproxy composite/ build:
curl -X POST "http://localhost:7081/api/tasks/reload-entities?types=services&ids=*"
— before: warning + duplicate-key exception, external apiCatalogAdditionalEntries
listed multiple times. After: clean reload, no warnings, each entry listed once.