…ormalization

…, ring buffer, config)

…in delegation

- Fix include order in log_manager.hpp: system/external before project headers
- Document OR-aggregation semantics in LogProvider::on_log_entry() doc comment
- Add thread safety comment to PluginManager::get_log_observers() explaining
  the pre-existing race condition with disable_plugin() (same as get_introspection_providers)
- Document dynamic_cast across dlopen boundary limitation (RTTI visibility hazard)
  with workaround guidance in load_plugins()
- Use named constants (Log::DEBUG/INFO/WARN/ERROR/FATAL) instead of magic numbers
  in level_to_severity() and severity_to_level()
- Use std::move(entry) in on_rosout() and inject_entry_for_testing() push_back calls

// @issue 208

Add LOGS to CapabilityBuilder::Capability enum and capability_to_name()
switch. Wire Cap::LOGS into Component and App discovery responses so that
entity detail endpoints advertise the /logs capability href.

// @issue 208

- LogHandlers: handle_get_logs, handle_get_logs_configuration,
  handle_put_logs_configuration. Components use prefix matching
  (all nodes under namespace); Apps use exact FQN matching.
- Wire LogManager into GatewayNode: get_log_manager() accessor,
  log_mgr_ member initialized after plugin_mgr_ setup so
  LogProvider plugins are registered before /rosout subscription starts.
- Register log_handlers.hpp in umbrella handlers.hpp and rest_server.hpp.
- Add test_log_handlers: 9 unit tests verifying 400 on missing matches.

…erver

// @issue 208

Register GET /components/{id}/logs, GET /apps/{id}/logs,
GET /{entity}/logs/configuration, and PUT /{entity}/logs/configuration
routes in RESTServer::setup_routes(). Instantiate LogHandlers alongside
other handler objects in RESTServer constructor.

// @issue 208

Parallel to get_raw(): sends PUT with JSON body and returns the raw
Response object without JSON parsing. Needed by the logging endpoint
integration tests to inspect responses from PUT /logs/configuration.

… endpoints

// @issue 208

Feature test covering:
- GET /apps/{id}/logs: returns 200 with items array, polls for log
  entries from startup, validates entry schema (id, timestamp,
  severity, message, context)
- GET /apps/{id}/logs?severity=error: severity filter excludes
  lower levels
- GET/PUT /apps/{id}/logs/configuration: config round-trip,
  invalid severity returns 400, max_entries=0 returns 400
- GET /components/{id}/logs and /logs/configuration: 200 responses
- Capability advertisement: 'logs' present in capabilities list
  for both components and apps

… traceability tags

…ceability tags

Fix RST definition list in docs/api/rest.rst (two consecutive terms sharing
a definition are not valid RST - combined each pair onto a single term line).

- Use extern "C" get_log_provider() query function in PluginLoader instead
  of dynamic_cast across the dlopen boundary (mirrors UpdateProvider /
  IntrospectionProvider discovery mechanism)
- Validate severity query parameter in GET /logs; return 400 for unknown values
- Validate max_entries type and range explicitly in PUT /logs/configuration;
  return 400 when present but not a positive integer
- Wrap LogProvider::on_log_entry() calls in try/catch to prevent plugin
  exceptions from crashing the ROS 2 subscription callback

…rocesses via ROS_DOMAIN_ID

The FaultManagerNode unit tests (test_fault_manager, test_sqlite_storage,
test_snapshot_capture, test_rosbag_capture) create in-process ROS 2 nodes on
the default DDS domain. The launch_testing integration tests (test_integration,
test_rosbag_integration) run concurrently and launch their own fault_manager_node
subprocess on the same domain, causing service calls to occasionally route to
the wrong server and making tests non-deterministically fail.

Fix: assign a dedicated ROS_DOMAIN_ID to each GTest binary that uses ROS 2 nodes
(62-65), keeping them isolated from integration test subprocesses that use the
default domain (0).

…efaults

- Add logs.buffer_size parameter (default: 200) to gateway_params.yaml
  and declare_parameter in GatewayNode; LogManager reads it at startup
- Reduce kDefaultBufferSize from 10 000 to 200 (ring buffer per node)
- Reduce LogConfig::max_entries default from 10 000 to 100 (GET /logs cap)
- Add Logging Endpoints section to README (quick list + detailed reference)
- Update test asserting the old max_entries default

- Remove REQ_INTEROP_062 (GET /logs/entries) and REQ_INTEROP_065
  (DELETE /logs/config) - neither exists in ISO 17978-3 §7.21
- Fix REQ_INTEROP_061 description: endpoint returns log entries, not an overview
- Fix REQ_INTEROP_063/064 URLs: /logs/config -> /logs/configuration
- Mark REQ_INTEROP_061/063/064 as verified (tests exist)
- Retag all @verifies REQ_INTEROP_062 -> REQ_INTEROP_061 in tests
- Update rest.rst: 10 000 -> 200 (buffer), 100 (max_entries default)
- Note 204 vs 200 deviation from SOVD spec for PUT /logs/configuration

… spec

Three integration test failures (jazzy, humble, rolling):

1. Log entries never appearing after startup:
   ROS 2 rosout messages carry logger names using '.' as separator
   (e.g. "powertrain.engine.temp_sensor") while entity FQNs use '/'.
   get_logs() now tries both slash-format and dot-format when matching
   buffer keys, so the default ring buffer resolves rosout entries
   correctly. Two unit tests added to cover this.

2. test_component_detail_includes_logs_capability failing:
   Component detail handler placed capabilities only under x-medkit,
   not at root level. Added response["capabilities"] at root level
   to match the app detail format (and the test expectation).

Remaining review comments addressed:
- All three 503 responses in log_handlers.cpp now use
  ERR_SERVICE_UNAVAILABLE (was ERR_INTERNAL_ERROR)
- log_handlers.hpp comment: "store/return" -> "return in query results"
- docs/api/rest.rst: "retain/return" -> "return in query results"

The two polling tests passed a condition that returned a plain bool
(True/False) instead of the response dict itself.  poll_endpoint_until
returns whatever the condition callable returns, so data ended up as
True and data['items'] raised TypeError: 'bool' object is not
subscriptable.

Change condition to lambda d: d if d.get('items') else None so the
full response dict is returned and data['items'] works correctly.