[BUG]序列化时子类型字段丢失

### 问题描述
json序列化JSON.toJSONString(object)，当bean中属性保存的实例是子类型时，可能引发序列化结果中子类型的属性丢失

### 环境信息

- 版本信息：[Fastjson2 2.0.48及以上]

### 重现步骤
@Getter
@Setter
public static class A {
private String fieldA;
}

@Getter
@Setter
public static class B extends A {
private String fieldB;
}

@Getter
@Setter
public static class Bean{
private A filed;
}

public static void main(String[] args) {

A a = new A();
a.setFieldA("aaa");
Bean bean1 = new Bean();
bean1.setFiled(a);

B b = new B();
b.setFieldA("aaa");
b.setFieldB("bbb");
Bean bean2 = new Bean();
bean2.setFiled(b);

// 此处输出{"filed":{"fieldA":"aaa"}}
System.out.println(com.alibaba.fastjson2.JSONObject.toJSONString(bean1));
// 此处输出{"filed":{"fieldA":"aaa"}}，更换toJsonString顺序则输出{"filed":{"fieldA":"aaa","fieldB":"bbb"}}
System.out.println(com.alibaba.fastjson2.JSONObject.toJSONString(bean2));
}

### 期待的正确结果
*对您期望发生的结果进行清晰简洁的描述。*

### 相关日志输出
*请复制并粘贴任何相关的日志输出。*

#### 附加信息
JSONFactory.defaultObjectWriterProvider中有个全局的cache用于保存ObjectWriter实例，当bean第一次序列化时，生成ObjectWriter实例并放入cache，后续序列化时使用缓存的ObjectWriter实例，对于这个bean的field也会生成对应的FieldWriterObject保存在ObjectWriter实例中，field序列化时会判断初始化的class和当前的class是否匹配，匹配就使用已有的FieldWriterObject，不匹配就生成新的FieldWriterObject

2.0.48的匹配逻辑增加了initValueClass.isAssignableFrom(valueClass)，导致如果从第二次开始存放子类型，使用的FieldWriterObject是父类型的，序列化结果中就会丢失子类型的属性
https://github.com/alibaba/fastjson2/commit/7d974f3024e5cebda4392b65c1e47f10cc239f70

Start with JSONFactory.defaultObjectWriterProvider and the FieldWriterObject matching logic described in the issue, then review commit 7d974f3024e5cebda4392b65c1e47f10cc239f70. Reproduce the supplied A, B, and Bean example with both serialization orders; done means the subtype fieldB is retained consistently.

Written by the indexing model from the issue text.