Location Manager: template-based locations, MongoDB persistence, and declarative node template registration - #228

- Fix LocationReservation.check() referencing removed start/end fields (now uses created/expires)
- Fix indentation bug in get_location_resources() making resource hierarchy query unreachable
- Fix Path.open() misuse in initialize() and add model_dump() for YAML serialization
- Fix dict indexing on Pydantic model in update_location() error message
- Fix add_location endpoint silently returning None on duplicate (now raises 409)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Update LocationClient methods (delete_location, attach_resource,
  detach_resource, get_location_resources) to use location_name
  matching the server's name-based routing
- Route get_location(location_id) through query endpoint to avoid
  path ambiguity with /location/{location_name}
- Remove redundant /location/id/{location_id} endpoint (covered by
  GET /location?location_id=...)
- Fix import_locations() to handle both list and dict YAML, close file handle
- Re-add reservation field to Location model for future reservation support
- Update Location docstring from "Definition" to "A location in the lab"
- Fix stale "by ID" docstrings across server, state handler, and client
- Update workcell_actions.py to use location_name for get_location_resources
- Update client and server tests to match new API surface
- Fix location manager initialization to load locations from file on
  startup (file is the source of truth for location definitions)
- Add depends_on to notebook_validator compose service (all managers + nodes)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Move persistent location data from Redis (ephemeral cache) to MongoDB
(document storage), keeping Redis for transient state only (locks, change
counters). This fixes the architectural mismatch where long-lived
configuration data was stored in an ephemeral store.

Key changes:
- LocationStateHandler: MongoDB for CRUD, Redis for transient state
- LocationManager: MongoDB version checking, seed file loading, bulk
  import/export endpoints, auto-migration from 0.7.1 Redis format
- LocationClient: import_locations POSTs to /locations/import endpoint,
  new export_locations and close methods
- LocationMigrator: one-time 0.7.1 Redis → MongoDB migration tool
- Settings: document_db_url, database_name, seed_locations_file (renamed
  from locations_file_path with backward-compat alias)
- Health: reports both document_db_connected and redis_connected
- LocalRunner: passes InMemoryMongoHandler to location manager
- 57 new tests (151 total location tests, 2723 full suite)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…te registration

- Add NodeResourceTemplateDefinition, NodeRepresentationTemplateDefinition,
  and NodeLocationTemplateDefinition types to node_types.py
- Add template_handler() method to AbstractNode that registers templates
  from ClassVar lists with per-template error isolation
- Call template_handler() before startup_handler() in _startup() so
  templates are available for create_resource_from_template() calls
- Move template registrations from startup_handler() to class-level
  declarations in liquidhandler, robotarm, and platereader example nodes
- Rename LocationRepresentationTemplate.schema to schema_def (with
  backward-compatible validation_alias) to avoid shadowing Python builtin
- Update LocationClient.init_representation_template() to accept both
  schema (deprecated) and schema_def parameters
- Add 17 tests covering registration, error isolation, call order,
  error logging, and definition type validation
- Use structured logging (kwargs) instead of f-strings in all new log calls

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

# Conflicts:
#	examples/example_lab/compose.yaml
#	src/madsci_common/madsci/common/document_db_migration_tool.py
#	src/madsci_common/madsci/common/document_db_version_checker.py
#	src/madsci_common/madsci/common/local_backends/local_runner.py
#	src/madsci_common/tests/test_mongodb_version_checker.py

…r-neutral terms

Systematically update all remaining references to proprietary database
products (MongoDB, Redis, MinIO) with vendor-neutral terminology
(document database, cache, object storage) or FOSS product names
(FerretDB, Valkey, SeaweedFS) throughout comments, docstrings, config
fields, documentation, and test files.

Breaking changes (beta):
- env_prefix MONGODB_ → DOCUMENT_DB_ (DocumentDBBackupSettings)
- env_prefix MONGODB_MIGRATION_ → DOCUMENT_DB_MIGRATION_ (DocumentDBMigrationSettings)
- Fields redis_host/port/password → cache_host/port/password (WorkcellManagerSettings, LocationManagerSettings)
- Fields redis_connected → cache_connected (health models)
- Docker types REDIS_PORT → CACHE_PORT
- Backup metadata backup_type value "mongodb" → "document_db"

All old field names remain accepted via validation_alias for backward
compatibility. 2889 tests passing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Continues the FOSS terminology audit by renaming the cache handler
abstraction layer from Redis-specific names to generic cache names,
consistent with the earlier config field renames (redis_host → cache_host).

Renames:
- redis_handler.py → cache_handler.py
- RedisHandler → CacheHandler (ABC)
- PyRedisHandler → PyCacheHandler (real implementation)
- InMemoryRedisHandler → InMemoryCacheHandler (test implementation)
- redis_handler param/attr → cache_handler/_cache_handler

Wire protocol references (redis.Redis, pottery.RedisDict, InMemoryRedisClient)
are kept as-is since they refer to the underlying client libraries.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

… manager index name conflicts

The version checker's validate_or_fail() only auto-initialized completely empty
databases (current=None), but refused to start when collections existed without
version tracking (current=0.0.0). This caused managers to crash-loop when
FerretDB volumes persisted across restarts. Extended auto-initialization to also
handle the 0.0.0 case, since both mean no prior version tracking existed.

Also fixed LocationStateHandler creating indexes with names that conflicted with
schema.json definitions (template_name_unique vs repr_template_name_unique /
loc_template_name_unique), causing FerretDB IndexOptionsConflict errors.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Use timezone-aware datetime.now(timezone.utc) in location types and server
- Move state_change_marker and shutdown from class to instance attributes
- Add efficient count methods to LocationStateHandler for health endpoint
- Defer reconciliation in init_representation_template to background loop
- Fix representations guard to handle both None (legacy) and empty dict
- Remove bogus experiment_id_1 unique index from experiment manager schema
  (experiment_id is aliased to _id by to_mongo(), so the index on the
  non-existent field caused DuplicateKeyError on second insert)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>