fix: cluster token checker

[leeyazhou]

1. ClusterFlowChecker.acquireClusterToken
2. ClusterParamFlowChecker.acquireClusterToken

Describe what this PR does / why we need it

集群限流情况下, windowIntervalMs大于1000ms时, 会出现限流不准确的情况.

测试场景:

FlowRule{count =2, clusterMode=true, clusterConfig { windowIntervalMs: 3000}}

        double latestAvg = metric.getAvg(ClusterFlowEvent.PASS);
        double globalThreshold = calcGlobalThreshold(rule) * ClusterServerConfigManager.getExceedCount();
        double nextRemaining = globalThreshold - latestAvg - acquireCount;

        if (nextRemaining >= 0) {
            return new TokenResult(TokenResultStatus.OK)
                .setRemaining((int) nextRemaining)
                .setWaitInMs(0);
        }

代码执行过程:

1. 10:00:01.100, latestAvg=0, nextRemaining (1) >=0 通过
2. 10:00:01.200, latestAvg=0.333, nextRemaining (0.66) >=0 通过
3. 10:00:01.300, latestAvg=0.666, nextRemaining (0.33) >=0 通过
4. 10:00:01.400, latestAvg=1, nextRemaining (0) >=0 通过

限流规则本来是3s通过2个, 现在已经通过了4个, 说明这个限流规则不准确

Does this pull request fix one issue?

Describe how you did it

ClusterFlowChecker.acquireClusterToken, 获取时间窗口内已通过的请求数时取总数, 而不是取QPS

        double latestSum = metric.getSum(ClusterFlowEvent.PASS);
        double globalThreshold = calcGlobalThreshold(rule) * ClusterServerConfigManager.getExceedCount();
        double nextRemaining = globalThreshold - latestSum - acquireCount;

        if (nextRemaining >= 0) {
            return new TokenResult(TokenResultStatus.OK)
                .setRemaining((int) nextRemaining)
                .setWaitInMs(0);
        }

Describe how to verify it

Special notes for reviews

[CLAassistant]

All committers have signed the CLA.