state sync: improve behavior if epoch ends during state sync

[saketh-are]

Description

When a new node joins the network, it is possible for the epoch to end while the node is still in state sync.

In the current implementation:

• The syncing nodes does not track the head of the chain and will not detect that the epoch has ended. It will continue to send requests for parts of the older snapshot it is trying to obtain.
• The peers in the network hosting the snapshot will delete it as soon as the epoch ends.

This is not currently a critical issue because the syncing node can finish state sync from the cloud. However, we should still fix it to reduce dependence on the centralized storage. We can take two approaches:

• Increase the retention period for peer-hosted snapshots so that they are still available after the epoch ends.
• Have the syncing node detect that the epoch has ended, at which point it would start a fresh state sync to the beginning of the new epoch.