feat: send drop decisions in batch

[VinozzZ]

Which problem is this PR solving?

To reduce network traffic from nodes communicating about trace decisions, drop decisions will be sent in batches. The batching logic is based on either a maximum batch size or a timer. Refinery will send a batch when the batch size limit is reached or when the timer expires, whichever comes first.

Short description of the changes

• add sendDropDecisions goroutine
• add two new config options for the batch size and the ticker interval. This is mostly for easier testing. We may not include these two configs in the final release.

[kentquirk]

I'm happy with this, but left one comment to think about.

[kentquirk]

This gives us a possible situation like this:

• Max size is reached at time T
• We send the batch
• We get one decision
• At T+1ms the ticker fires
• We send that batch

If the number of decisions per second is close to the max batch size, then the above is very likely to happen.

I've thought of two possible ways to deal with this:

• Keep a timestamp of the oldest item in the batch; if it's less than our ticker time, then don't send yet.
• Don't use a ticker, use a timer, and restart the timer after every send, no matter why we sent it.

What do you think?

[VinozzZ]

I like the timer idea. I will update the code to do that