Quality upgrade for NSM chain element testing

[denis-tingaikin]

Motivation 

Most chain elements are handling resources somehow. Some elements create a private map for storing states; some of them use a metadata map from the context; some of them handle context-related events; and some of them do not, some of them doesn't handle error on the refresh request.

The inconsistent behavior produces memory leaks and unexpected behavior.

It's a root cause of networkservicemesh/cmd-nsmgr#675.

Solution

Define a set of must have checks for the chain element that will check does the element follow required conditions. The interface could be simple:

package sandbox

func TestClientEdgeCases(t *testing.T, c networkservice.NetworkServiceClient) {/* TODO  */ }
func TestServerEdgeCases(t *testing.T, s networkservice.NetworkServiceServer) {/* TODO */ }

Example of usage:

package metrics_test

import "sandbox"
import "metrics"

// should be added in each chain element
func TestEdgeCases(t *testing.T) {
    sandbox.TestServerEdgeCases(t, metrics.NewServer())
    sandbox.TestClientEdgeCases(t, metrics.NewClient())
}

Scenarios

1. sequence of request + close => should not produce cpu/memory leaks
2. sequence of request + request failed by timeout => should not produce cpu/memory leaks
3. sequence of request + request failed by error in next => should not produce cpu/memory leaks
4. sequence of request + error in next => should not produce cpu/memory leaks
5. sequence of request + error in next + close => should not produce cpu/memory leaks