retries设置问题

- [ ] I have searched the [issues](https://github.com/apache/dubbo/issues) of this repository and believe that this is not a duplicate.

### Environment

* Dubbo version: 3.1.0
* Operating System version: centos 7.9、windows10
* Java version: 8、11

### Steps to reproduce this issue

1.application.yaml配置如下：
# 提供者服务配置
provider:
# 默认快速失败
cluster: failfast
# 不进行重试
retries: 0
消费者调用超时，没有进行快速失败，并且进行了重试三次
2.@DubboService(cluster = "failfast",methods = {@Method(name = "test", retries = 0)})
即使在提供者针对方法级别进行配置，还是和第一个问题产生的结果一致

Start by reproducing the timeout scenario from application.yaml with provider cluster failfast and retries: 0, then compare it with the @DubboService and @Method configuration shown. Trace how these settings are resolved for the consumer call; done means the retry behavior matches the documented configuration or the discrepancy is fixed and verified.

Written by the indexing model from the issue text.