refactor: modularize yaml2pptx into layered architecture

Refactor yaml2pptx.py from a 1,245-line monolithic script into a modular
architecture with clear separation of concerns. The entry point is now
127 lines, with business logic distributed across focused modules.

Architecture:
- core/: Domain models (elements, template, presentation)
- loaders/: YAML loading and validation
- renderers/: PPTX and HTML rendering
- preview/: Flask preview server
- utils.py: Shared utilities

Key improvements:
- Element abstraction layer using dataclass with validation
- Renderer logic built into generator classes
- Single-direction dependencies (no circular imports)
- Each module 150-300 lines for better readability
- Backward compatible CLI interface

Documentation:
- README.md: User-facing usage guide
- README_DEV.md: Developer documentation

OpenSpec:
- Archive refactor-yaml2pptx-modular change (63/70 tasks complete)
- Sync 5 delta specs to main specs (2 new + 3 updated)

core/presentation.py

@@ -0,0 +1,91 @@
+"""
+演示文稿模块
+
+管理整个演示文稿的生成流程。
+"""
+
+from pathlib import Path
+from loaders.yaml_loader import load_yaml_file, validate_presentation_yaml
+from core.template import Template
+from core.elements import create_element
+
+
+class Presentation:
+    """演示文稿类，管理整个演示文稿的生成流程"""
+
+    def __init__(self, pres_file, templates_dir=None):
+        """
+        初始化演示文稿
+
+        Args:
+            pres_file: 演示文稿 YAML 文件路径
+            templates_dir: 模板目录
+        """
+        self.pres_file = Path(pres_file)
+        self.templates_dir = templates_dir
+
+        # 加载演示文稿文件
+        self.data = load_yaml_file(pres_file)
+        validate_presentation_yaml(self.data, str(pres_file))
+
+        # 获取演示文稿尺寸
+        metadata = self.data.get('metadata', {})
+        self.size = metadata.get('size', '16:9')
+
+        # 模板缓存
+        self.template_cache = {}
+
+    def get_template(self, template_name):
+        """
+        获取模板（带缓存）
+
+        Args:
+            template_name: 模板名称
+
+        Returns:
+            Template 对象
+        """
+        if template_name not in self.template_cache:
+            self.template_cache[template_name] = Template(
+                template_name, self.templates_dir
+            )
+        return self.template_cache[template_name]
+
+    def render_slide(self, slide_data):
+        """
+        渲染单个幻灯片
+
+        Args:
+            slide_data: 幻灯片数据字典
+
+        Returns:
+            dict: 包含 background 和 elements 的字典
+        """
+        if 'template' in slide_data:
+            # 使用模板
+            template_name = slide_data['template']
+            template = self.get_template(template_name)
+            vars_values = slide_data.get('vars', {})
+            elements = template.render(vars_values)
+
+            # 合并背景（如果有）
+            background = slide_data.get('background', None)
+
+            # 将元素字典转换为元素对象
+            element_objects = [create_element(elem) for elem in elements]
+
+            return {
+                'background': background,
+                'elements': element_objects
+            }
+        else:
+            # 自定义幻灯片
+            elements = slide_data.get('elements', [])
+
+            # 将元素字典转换为元素对象
+            element_objects = [create_element(elem) for elem in elements]
+
+            return {
+                'background': slide_data.get('background'),
+                'elements': element_objects
+            }