Create CRON to purge S3 buckets based on commit hash

[willmcvay]

Background context or User story:

We get occasional chunk missing errors in dev when we deploy because currently we replace the content of a bucket with the new deployment. If a user has a session, the script will be missing when they navigate. eg of this behaviour https://sentry.io/organizations/reapit-ltd/issues/1623235951/?project=1885603&referrer=slack

Specification or Acceptance Criteria:

• We should support at least 2 recent versions of each app in an S3 bucket.
• We should update our webpack scripts so the master HEAD commit hash is prepended to the output script file names. This means we can identify the scripts by release
• We should have a CRON job that runs weekly to purge all S3 buckets of all but the last two releases by most recent commit hash.

[phmngocnghia]

We have dicussed about using service-worker, offline-plugin API to check if the assets have been changed to do a refetched-navigation and agreed to try this method: https://zach.codes/handling-client-side-app-updates-with-service-workers/

The service worker implemented by the lib “offline-plugin” (we are currently using it) allows us to detect when the new code is out by checking the status of the manifest file automatically every X time, or checking it manually, then we can use that info to do a re-fetched navigation when the user changes their route

[willmcvay]

• Had to remove S3 Cron file, this needs re-instating.

Also;

• When a new release is available and I click on the button, throws this error: https://sentry.io/organizations/reapit-ltd/issues/1636945628/?referrer=slack
• When I have clicked on the new release button, the toast should submit. Doesn't currently in dev which could be linked to the error