tick延时有点大，spdlog显示从ctp到策略达500us about ft

[xywanz]

spdlog本身耗时就很大，会有一两百微秒的时延，所以用spdlog测量的时间误差比较大
先保证时间能够准确测量，在tick中加入当前毫秒时间字段
struct TickData : public pubsub::Serializable {
uint64_t current_time_us; // 加入该字段
MarketDataSource source;
datetime::Datetime datetime;

uint32_t ticker_id;
double last_price = 0;
double open_price = 0;
double highest_price = 0;
double lowest_price = 0;
double pre_close_price = 0;
double upper_limit_price = 0;
double lower_limit_price = 0;
uint64_t volume = 0;
uint64_t turnover = 0;
uint64_t open_interest = 0;

double ask[kMaxMarketLevel]{0};
double bid[kMaxMarketLevel]{0};
int ask_volume[kMaxMarketLevel]{0};
int bid_volume[kMaxMarketLevel]{0};

struct {
double iopv;
} etf;

SERIALIZABLE_FIELDS(current_time_us, source, datetime, ticker_id, last_price, open_price, highest_price,
lowest_price, pre_close_price, upper_limit_price, lower_limit_price, volume,
turnover, open_interest, ask, bid, ask_volume, bid_volume, etf.iopv);
};

在收到行情时，先在tick中记录当前时间
uint64_t GetCurrentTimeUs() {
timespec ts;
clock_gettime(CLOCK_MONOTONIC, &ts);
return ts.tv_nsec / 1000 + ts.tv_sec * 1000000;
}

tick.current_time_us = GetCurrentTimeUs();

在策略端收到行情时
uint64_t current_us = GetCurrentTimeUs();
uint64_t delay_us = current_us - tick.current_time_us;

tick延迟大应该是由于ZeroMQ延迟大。
解决办法：开发基于共享内存的行情分发模块，可以让通信时延降低至1us以内。