Add project Schedule page with interactive Gantt chart

Adds a Schedule tab (Kitsu-style production schedule) under each
project: tasks grouped by type with drag-to-reschedule bars, Day/Week/
Month zoom, manual date-range control, weekend shading, a frozen task
column, and a two-tier month/date axis header. Requires a new
start_date field on Task (start_date was previously missing; only
deadline existed) and shadcn DatePicker inputs replace native date
inputs on the Schedule toolbar and TaskDetailPanel's Start Date/
Deadline fields.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

backend/generate_schedule_example_data.py

@@ -0,0 +1,121 @@
+#!/usr/bin/env python3
+"""
+One-off script to generate example start_date/deadline data on shot tasks,
+for testing/demoing the Project Schedule (Gantt) page.
+
+Reuses the same shot-sampling approach as generate_note_example_data.py: the
+dev database is heavily populated with disposable load-test fixtures, so this
+picks a small, bounded sample of shots (from Dragon Quest) that have the
+normal 4-task set, rather than scheduling all ~1400 shot tasks.
+
+For each sampled shot, tasks are scheduled in pipeline order (layout ->
+animation -> simulation -> lighting -> compositing, any other task types
+appended after) with realistic overlapping durations, staggered around
+today's date so the chart's "Today" marker falls inside the visible range.
+
+Idempotent: skips any task that already has a start_date set, so it's safe
+to re-run without clobbering manually-edited dates.
+"""
+
+from datetime import date, timedelta
+from sqlalchemy import create_engine
+from sqlalchemy.orm import sessionmaker
+from database import DATABASE_URL, Base
+from models.shot import Shot
+from models.task import Task
+from models.project import Project
+import logging
+import random
+
+import models  # noqa: F401 - ensures every model is registered on Base.metadata
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+PROJECT_NAME = "Dragon Quest"
+SAMPLE_SIZE = 15
+RNG_SEED = 42
+
+# Canonical pipeline order used to sequence a shot's tasks; anything not
+# listed here keeps its natural (query) order, appended at the end.
+PIPELINE_ORDER = ["layout", "animation", "simulation", "lighting", "compositing", "previz"]
+
+
+def pipeline_index(task_type: str) -> int:
+    return PIPELINE_ORDER.index(task_type) if task_type in PIPELINE_ORDER else len(PIPELINE_ORDER)
+
+
+def generate_schedule_example_data():
+    engine = create_engine(DATABASE_URL)
+    SessionLocal = sessionmaker(autocommit=False, autoflush=False, bind=engine)
+    Base.metadata.create_all(bind=engine)
+
+    db = SessionLocal()
+    rng = random.Random(RNG_SEED)
+    today = date.today()
+
+    try:
+        project = db.query(Project).filter(Project.name == PROJECT_NAME).first()
+        if not project:
+            logger.error(f"Project '{PROJECT_NAME}' not found")
+            return
+        logger.info(f"Project '{PROJECT_NAME}' (id={project.id})")
+
+        # Shots with the normal 4-task set (excludes shots with zero tasks
+        # and the handful of outliers with extra bulk-test tasks piled on).
+        shots = db.query(Shot).filter(
+            Shot.project_id == project.id, Shot.deleted_at.is_(None)
+        ).order_by(Shot.id).all()
+        candidates = [
+            s for s in shots
+            if db.query(Task).filter(Task.shot_id == s.id, Task.deleted_at.is_(None)).count() == 4
+        ]
+        sample_shots = candidates[:SAMPLE_SIZE]
+        logger.info(f"{len(candidates)} candidate shots with a 4-task set, sampling {len(sample_shots)}")
+
+        tasks_scheduled = 0
+        tasks_skipped = 0
+
+        for shot in sample_shots:
+            tasks = db.query(Task).filter(
+                Task.shot_id == shot.id, Task.deleted_at.is_(None)
+            ).all()
+            tasks.sort(key=lambda t: pipeline_index(t.task_type))
+
+            # Stagger each shot's pipeline start across a window that straddles
+            # today, so the demo chart shows a mix of past/current/future work.
+            shot_start = today + timedelta(days=rng.randint(-20, 20))
+            cursor = shot_start
+
+            for task in tasks:
+                if task.start_date is not None:
+                    tasks_skipped += 1
+                    # Still advance the cursor so later tasks in this shot
+                    # don't bunch up if an earlier one was already scheduled.
+                    cursor = task.deadline or cursor
+                    continue
+
+                duration = rng.randint(4, 12)
+                task.start_date = cursor
+                task.deadline = cursor + timedelta(days=duration)
+                tasks_scheduled += 1
+
+                # Next task starts a little before this one ends (realistic
+                # pipeline overlap) rather than strictly back-to-back.
+                cursor = task.deadline - timedelta(days=rng.randint(0, 4))
+
+        db.commit()
+        logger.info(f"Tasks scheduled: {tasks_scheduled}, skipped (already scheduled): {tasks_skipped}")
+
+    except Exception as e:
+        logger.error(f"Generation failed: {e}")
+        db.rollback()
+        raise
+    finally:
+        db.close()
+
+
+if __name__ == "__main__":
+    logger.info("Generating schedule example data...")
+    generate_schedule_example_data()
+    logger.info("Done!")