[1.0.3] Fix flaky nodeos_snapshot_diff_test due to a race condition

[linh2931]

Recently, nodeos_snapshot_diff_test fails in almost every CICD run.

This is due to a race condition that at the startup of head mode node, net thread and http thread can start to work in different order. If http thread processes schedule_snapshot request after net thread starts to sync with the irreversible mode node, schedule_snapshot request will miss the scheduled block number. If it is before net thread starts to sync with the irreversible mode node, schedule_snapshot request will catch the scheduled block number and the snapshot is taken.

This PR changes to start up the head mode node while the irreversible mode node is down so that net thread will not sync, make schedule_snapshot request, and finally restart irreversible node.

Test run in CICD for 3 times. No failure happened.

Resolves #846

[ericpassmore]

Note:start
category: Tests
component: Snapshot
summary: Fix flaky nodeos_snapshot_diff_test due to a race condition.
Note:end