Remove resetting Auth credentials in reverse-proxy

[akatashev]

Related issue: #979

Description

This PR removes setting Auth credentials in reverse-proxy.
@andrein was right about the fact that lines

orchestrator/go/http/raft_reverse_proxy.go

Lines 42 to 45 in d536bc6

	switch strings.ToLower(config.Config.AuthenticationMethod) {
	case "basic", "multi":
	r.SetBasicAuth(config.Config.HTTPAuthUser, config.Config.HTTPAuthPassword)
	}

reset auth credentials for every request going through reverse-proxy and therefore provide this request read/write permissions even for a readonly user.

These lines are unnecessary, because there are 3 basic scenarios:

1. Username and password are incorrect, so an attempt to sent a request receives 401 Unauthorized. In this case we shouldn't care about proxying, because it never reaches the proxy.
2. Leader and follower have different configuration. For example, a follower has "AuthenticationMethod": "multi" and a leader has "AuthenticationMethod": "basic". But that isn't a proxy problem.
3. Leader and follower are configured correctly, so if Authorization in a request is good for a follower, it should be good for a leader when request is proxied to it. So it means there is no need to modify authorisation for a request.

Testing

During the whole testing both servers had the following configuration:

{
  "AuthenticationMethod": "multi",
  "HTTPAuthUser":         "dba_team",
  "HTTPAuthPassword":     "time_for_dinner"
}

Before fix behaviour (Orchestrator version built from the latest master state):

bash-4.2$ curl https://orch-leader:3000/api/forget-cluster/mysql-cluster -k -u readonly:123
{"Code":"ERROR","Message":"Unauthorized","Details":null}
bash-4.2$ curl https://orch-follower:3000/api/forget-cluster/mysql-cluster -k -u readonly:123
{"Code":"OK","Message":"Cluster forgotten: mysql-server-1:3306","Details":null}

Request to the leader is rejected, because it's executed as readonly, but the same request sent to the follower is accepted, because its user is implicitly changed to dba_team by reverse-proxy.

After fix behaviour:

bash-4.2$ curl -k https://orch-leader:3000/api/forget-cluster/mysql-cluster -u readonly:123
{"Code":"ERROR","Message":"Unauthorized","Details":null}
bash-4.2$ curl -k https://orch-follower:3000/api/forget-cluster/mysql-cluster -u readonly:123
{"Code":"ERROR","Message":"Unauthorized","Details":null}

Since credentials are not rewritten by reverse-proxy, both requests produce expected results.

To ensure that we didn't break actual functionality, let's execute a valid request through a follower:

bash-4.2$ curl -k https://orch-follower:3000/api/forget-cluster/mysql-cluster -u dba_team:time_for_dinner
{"Code":"OK","Message":"Cluster forgotten: mysql-server-1:3306","Details":null}

It shows that the request was relayed and executed successfully.

[shlomi-noach]

Thank you for the fix and for the analysis!