Separate responsibilities during check out process

At the moment, the check out process has the following responsibilities:

• interpreting the request from the client
• Fetching the relevant information to process the check out
• applying the check out business logic
• persisting any changes in storage
• sending notices
• publishing messages for the circulation log etc
• responding to the client

These are all mingled together in a single flow. This makes the code harder to understand and easier to accidentally break. A recent example of this was CIRC-995 which introduces business logic that should fail a check out after changes to item had already been persisted