Improve backoff strategy for redis reconnection

[awarus]

Sometimes we may escalate the problem of redis reconnection because of re-run pod every time: https://neondb.slack.com/archives/C07R2JP05CJ/p1741259453159759, we need to improve that