fix(redis): treat ignorable Redis errors as warnings

[aramissennyeydd]

Summary

We've been seeing intermittent CI failures caused by messages like,

==[ e (build) ]==================================================[ 10 of 10 ]==
node:events:496
      throw er; // Unhandled 'error' event
      ^

SocketClosedUnexpectedlyError: Socket closed unexpectedly

We know that the underlying library @redis/client handles retries but it looks like there is no .on('error' handler to listen for retryable errors so it is crashing the process. This PR addresses that by adding an .on('error' handler and configuring the retry strategy to give up after a couple of failed connection attempts (I chose 5 as a good number). This also necessitated bumping the package version to latest.

Details

Added an 'error' handler to prevent the entire process from crashing due to unhandled errors. This is hidden in the docs, https://github.com/redis/node-redis#events but is definitely causing issues for us.

How it was tested

Ran the rush-redis-cobuild-integration-test locally and docker-compose down-ed the redis server halfway through a run. Before this change, the Rush process would immediately crash - with this change, if you docker-compose up -d again within the retryStrategy timeout, the process continues until success. If you keep the redis server off, the process will die with a message like,

--[ FAILURE: b (build) ]-------------------------------------[ 2.09 seconds ]--

Invoking: node ../build.js b 
start b
done
Failed to set completed_state(cobuild:completed:foo:56ea08cd70a55cacbb53405171864ebd5cf2ad0b)_package(b)_phase(_phase:build): 


Operations failed.

rush cobuild (11.97 seconds)

ERROR: Failed to disconnect to redis server: The client is closed

Impacted documentation

[aramissennyeydd]

@iclanton Is this good to merge?