[BUG] Synthetic components return empty results from /logs and /bulk-data resource collections

[bburda]

Bug report

Steps to reproduce

1. Launch the gateway with at least one runtime-discovered ROS 2 node grouped into a synthetic / runtime-discovered component (any node whose component is not declared in a manifest):

   ros2 launch ros2_medkit_gateway gateway.launch.py
   ros2 launch ros2_medkit_integration_tests demo_nodes.launch.py
   

2. Identify the synthetic component that hosts a demo app:

   curl -s localhost:8080/api/v1/components | jq '.items[].id'
   COMP_ID=...  # the one whose /hosts contains a demo app
   curl -s localhost:8080/api/v1/components/$COMP_ID/hosts
   

3. Query log entries for that component:

   curl -s "localhost:8080/api/v1/components/$COMP_ID/logs?severity=debug" | jq
   

4. Query bulk-data rosbag descriptors for the same component (after a fault triggers a capture):

   curl -s "localhost:8080/api/v1/components/$COMP_ID/bulk-data/rosbags" | jq
   

Expected behavior

Both endpoints return entries aggregated from the apps hosted by the component, with x-medkit.aggregation_level=component and a non-empty items list (mirroring how /areas/{id}/logs and /functions/{id}/logs already aggregate).

Actual behavior

Both endpoints return {"items": []} even when the component groups N apps that have logs / rosbags. No error, no warning - just a quiet empty response.

Root cause is the same in both handlers: synthetic / runtime-discovered components have an empty fqn and an empty namespace_path, while manifest-defined components have non-empty values.

• LogHandlers::handle_get_logs calls LogManager::get_logs({entity.fqn}, prefix_match=true, ...) for the COMPONENT branch. With an empty fqn, the prefix-match against per-node logger name buffers selects nothing and returns zero items. AREA and FUNCTION already aggregate via hosted apps' effective_fqn(), so they work correctly; COMPONENT is the only entity type still on the legacy path.
• BulkDataHandlers::get_source_filters has a special case for FUNCTION (cache.get_apps_for_function() + effective_fqn()) but falls through to entity.fqn.empty() ? entity.namespace_path : entity.fqn for COMPONENT. Both fields empty → returns {} source filters, so handle_list_descriptors (rosbag listing) and the ownership check in handle_download get nothing to work with.

Environment

• ros2_medkit version: main (reproduces on every recent release; introduced when synthetic components became normal in Rethink runtime discovery entity mapping to align with SOVD entity model #325)
• ROS 2 distro: any (Jazzy / Humble / Rolling)
• OS: any

Additional information

Fix mirrors the AREA / FUNCTION pattern already in both handlers - resolve hosted apps via cache.get_apps_for_component(), build their effective_fqn() list, fall through to the namespace-prefix path only when the component has no hosted apps but declares a non-empty namespace (manifest deployments where a component groups topics rather than nodes).

A related but distinct symptom on the per-entity faults handler is tracked separately - see comment on #380 for details. That one is a scope-leak rather than an empty-result bug and needs a different fix (changing the GetFault service contract), so it is not in scope here.