refactor: optimize perf-counter

[foreverneverer]

The perf-counter has some problem:

• Existed many redundant counter, such as *.cps-15min,latency.p50 etc, this pr delete these.
• No existed suitable tag: user usually don't set the tag config or we don‘t get it though it is set, this pr will add first ip of meta list and table name as default tag which also join to the user tag if it is set. add meta=ip tag is because sometimes different meta_server has same table name, so that if you search table=name , we will not distinguish it. The follow is an tag example:

user tag string: cluster=xiaomi
the final tag : cluster=xiaomi,meta=ip,table=name

• enable_perf_counter default value is set true

[acelyc111]

this pr will add first ip of meta list and table name as default tag which also join to the user tag if it is set.

cluster meta list is mutable, how to deal with the case when meta server changed?

[foreverneverer]

this pr will add first ip of meta list and table name as default tag which also join to the user tag if it is set.

cluster meta list is mutable, how to deal with the case when meta server changed?

Yeah, I have considered it. Actually, only table tag is included in different clusters, meta ip may be used when we search. Now, since meta list is not host name, I haven't find a way the same as sever client to add invariable tag for one cluster.

[foreverneverer]

New commit has delete meta ip

.

[foreverneverer]

I will re-test it manually before merged