feat(pii): NER tier engine — privacy-filter.cpp backend + NER-centric PII filter (#10360)

Squashed feat/pii-ner-tier-engine rebased onto master (was 45 commits; see
backup/pii-ner-tier-engine-prerebase). Net change:

- privacy-filter.cpp: standalone GGML engine for the openai-privacy-filter
  PII/NER token classifier, wired as a LocalAI gRPC backend (CPU/CUDA/Vulkan).
  TokenClassify moves off the patched llama.cpp path onto this backend.
- PII filter reworked to be NER-centric (encoder/NER detection tier scanning
  whole conversations as one document), with a recreated bounded restricted-
  regex secret-matching pattern detector tier alongside it (per-model
  pii_detection.builtins / .patterns + core/services/routing/piipattern).
- Detection labelled by source (ner vs pattern); backend trace / confidence /
  debug observability; analyze/redact exposed as a synchronous API.
- Instance-wide default detector policy + per-usecase default-on; request
  filtering extended to completions, embeddings, edits & Ollama.
- React UI: NER-centric PII editor, detector-models table, pattern/builtins
  editor, middleware default-policy UI.
- Gallery: privacy-filter-multilingual token-classify model + NER install
  filter; token_classify known_usecase; batch sized to context for NER models.
  privacy-filter backend registered in the backend gallery (cpu/vulkan/cuda-13
  meta + image entries with a capabilities map) matching its CI matrix jobs,
  and an /import-model auto-detect importer (PrivacyFilterImporter, narrow
  privacy-filter GGUF detection) replacing the prior pref-only registration.

Reconciled against master's independent evolution:

- Dropped master's PIIPatternOverrides feature (global-pattern runtime
  overrides + /api/pii/patterns API + runtime_settings.json persistence). The
  per-model NER + pattern-detector design supersedes it; it was built on the
  global redactor pattern set this branch replaced.
- Reverted the llama.cpp Score carry-patch (0006-server-task-type-score):
  removed the patch and restored master's grpc-server.cpp Score RPC (direct
  llama_decode, slot-loop bypass) and LLAMA_VERSION pin, plus master's
  model_config validation forbidding score + chat/completion/embeddings on
  llama-cpp. token_classify is unaffected (it runs on the privacy-filter
  backend, not llama-cpp).

Assisted-by: Claude:claude-opus-4-8 [Claude Code]

Signed-off-by: Richard Palethorpe <io@richiejp.com>

core/http/middleware/route_model.go

@@ -391,18 +391,12 @@ func buildClassifier(cfg *config.ModelConfig, deps ClassifierDeps) (router.Class
 }
 
 // assertClassifierDeclaresScore refuses to build the score classifier
-// unless classifier_model's config declares FLAG_SCORE. The actual
-// usecase-conflict check (score + chat/completion/embeddings on
-// llama-cpp) lives in ModelConfig.Validate() and fires at config load
-// and save time — by the time we get here, any model that reached the
-// loader is already conflict-free. This check just refuses to bind a
-// model that never declared itself for Score in the first place; that
-// model could be a misconfigured chat model the operator pointed at
-// by accident, and without FLAG_SCORE the validator never saw it.
+// unless classifier_model's config declares FLAG_SCORE. This check only
+// refuses to bind a model that never declared itself for Score in the
+// first place; that model could be a misconfigured chat model the
+// operator pointed at by accident.
 //
-// When lookup is nil (test wiring) the check is skipped and we fall
-// back to the C++ backend's runtime tripwire as the last line of
-// defence.
+// When lookup is nil (test wiring) the check is skipped.
 func assertClassifierDeclaresScore(classifierModel string, lookup ModelConfigLookup) error {
 	if lookup == nil {
 		return nil
@@ -416,8 +410,8 @@ func assertClassifierDeclaresScore(classifierModel string, lookup ModelConfigLoo
 	if !cfg.HasUsecases(config.FLAG_SCORE) {
 		return fmt.Errorf(
 			"router classifier score: classifier_model %q does not declare the "+
-				"score usecase. Add `known_usecases: [score]` to its config so "+
-				"the loader can reject conflicting usecase combinations",
+				"score usecase. Add `known_usecases: [score]` (alongside any other "+
+				"usecases the model serves) to its config",
 			classifierModel)
 	}
 	return nil