Proposal: combine client metrics into server and add options

[rogerogers]

• The current code needs to create an http server to monitor the exposed metrics interface every time the middleware is used. If an application uses multiple clients at the same time, many http servers need to be started.
• add options，e.g: buckets， go runtime metrics...

I've added some options on hertz-contrib/monitor-prometheus
I think we might save the registry the first we created, then reuse this， and register all metrics on it.

[rogerogers]

#9 maybe look like this.

[rogerogers]

• 现在的情况是每次调用中间件都需要创建一个http server去暴露指标给prometheus，如果一个服务创建了很多client，那么可能需要创建很多metric server。
• 添加一些选项，比如buckets，运行时指标