feat: frame/time pipeline split + output validation

- Add PipelineType enum + pipeline() to ProcessorType
- Split ProcessorPool into frame_slots (max 2) and time_slots (max 1)
- Add can_start_for() for pipeline-aware scheduling
- Add validate_output_file() — checks JSON validity before marking complete
- Add 3 unit tests for validate_output_file()
- Create DESIGN/FRAME_TIME_PIPELINE_V1.0.md (492 lines)

src/core/db/mod.rs

@@ -43,7 +43,7 @@ pub use mongodb_db::MongoDb;
 pub use postgres_db::{
     Bm25Result, CandidateRecord, CreateApiKeyConfig, FileIdentityRecord, FileRecord,
     HybridSearchResult, IdentityChunkRecord, IdentityDetailRecord, IdentityFaceRecord,
-    IdentityFileRecord, MonitorJob, MonitorJobStats, MonitorJobStatus, PostgresDb,
+    IdentityFileRecord, MonitorJob, MonitorJobStats, MonitorJobStatus, PipelineType, PostgresDb,
     ProcessorJobStatus, ProcessorResult, ProcessorType, ResourceRecord, VideoRecord, VideoStatus,
 };
 pub use qdrant_db::{QdrantDb, VectorPayload};

src/core/db/postgres_db.rs

@@ -559,6 +559,31 @@ impl ProcessorType {
             ProcessorType::FiveW1H,
         ]
     }
+
+    /// Pipeline type for scheduling: Frame-based, Time-based, or Cross (needs both).
+    pub fn pipeline(&self) -> PipelineType {
+        match self {
+            Self::Cut
+            | Self::Yolo
+            | Self::Face
+            | Self::Ocr
+            | Self::Pose
+            | Self::VisualChunk
+            | Self::Scene => PipelineType::Frame,
+
+            Self::Asr | Self::Asrx => PipelineType::Time,
+
+            Self::Story | Self::FiveW1H => PipelineType::Cross,
+        }
+    }
+}
+
+/// Pipeline classification for worker scheduling.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum PipelineType {
+    Frame,
+    Time,
+    Cross,
 }
 
 #[derive(Debug, Clone, Copy, Serialize, Deserialize, PartialEq)]