Coordinator: allow configuring volatile plan resolution

[snaury]

Volatile planning doesn't limit plan resolution, which results in plans up to 1000 times a second. While it's not inherently bad on its own, currently it causes an equivalent number of per-database broadcasts which is extremely expensive on large databases. Since sending packets to a large number of nodes takes a relatively long time, having such a high resolution actually harms tail latency as the result.

We need a way to limit plan tick resolution in these cases, e.g. to the standard 10ms. The default would stay at 1ms for improved latency in small databases.