Add optional blake3 hashing during asset scanning

- Make blake3 import lazy in hashing.py (only imported when needed)
- Add compute_hashes parameter to AssetSeeder.start(), build_asset_specs(), and seed_assets()
- Fix missing tag clearing: include is_missing states in sync when update_missing_tags=True
- Clear is_missing flag on cache states when files are restored with matching mtime/size
- Fix validation error serialization in routes.py (use json.loads(ve.json()))

Amp-Thread-ID: https://ampcode.com/threads/T-019c3614-56d4-74a8-a717-19922d6dbbee
Co-authored-by: Amp <amp@ampcode.com>

app/assets/services/bulk_ingest.py

@@ -36,6 +36,7 @@ class SeedAssetSpec(TypedDict):
     tags: list[str]
     fname: str
     metadata: ExtractedMetadata | None
+    hash: str | None
 
 
 class AssetRow(TypedDict):
@@ -163,7 +164,7 @@ def batch_insert_seed_assets(
         asset_rows.append(
             {
                 "id": asset_id,
-                "hash": None,
+                "hash": spec.get("hash"),
                 "size_bytes": spec["size_bytes"],
                 "mime_type": None,
                 "created_at": current_time,