Metric to track effectiveness of header recency estimation

We introduced the notion of header recency, which just checks if the header is recent via [isRecent](https://github.com/celestiaorg/go-header/blob/e50090545cc7e049d2f965d2b5c773eaa4a2c0b2/sync/sync_head.go#L192-L195) func. In case of the current subjective header is not recent, we forcefully re-request a new head. We need to know its effectiveness by checking if `blockTime+blocktime/2` is big enough not to cause unnecessary requests for the actual real head.

Start by reading sync/sync_head.go around the linked isRecent function and trace where a non-recent subjective header triggers a new head request. Define how effectiveness and unnecessary requests should be measured, then verify that the resulting metric supports evaluating the blockTime+blocktime/2 threshold.

Written by the indexing model from the issue text.