[Core] Support a list of multiple unique_keys in snapshots

[FishtownBuildBot]

At a minimum, update body to include a link to the page on docs.getdbt.com requiring updates and what part(s) of the page you would like to see updated.
Originating from this issue: dbt-labs/dbt-core#9992
Pull request: dbt-labs/dbt-core#10795

[runleonarun]

Documentation plan

• Update this page https://docs.getdbt.com/reference/resource-configs/unique_key
  • Revise the description to account for multiple unique keys
  • Add an example for using multiple unique keys: multiple id columns as unique key:

Use multiple unique keys

  ```
  snapshots:
    - name: orders_snapshot
      relation: source('jaffle_shop', 'orders')
      config:
        schema: snapshots
        unique_key: 
          - order_id
          - product_id
        strategy: timestamp
        updated_at: updated_at
  ```

• Update this page https://docs.getdbt.com/docs/build/snapshots#configuring-snapshots
  • Update the example in the table on this page to include an array for two unique ids [order_id, product_id]
  • Update description in table to say can be string or array.
• Possibly update examples in these pages to become arrays if it's recommended (in question 4)
  • https://docs.getdbt.com/docs/build/snapshots
  • https://docs.getdbt.com/reference/snapshot-properties
  • https://docs.getdbt.com/reference/commands/snapshot

Questions:

1. Does this plan seem OK / solid?
2. Does this replace concatenating columns as unique keys or is that still good practice?
3. Can we also use inline notation for these arrays For example, unique_key [order_id, product_id] or do we prefer block notations as shown in the example above?
4. Are multiple unique_keys backwards compatible to previous dbt versions or do I need to version these examples and content?
5. Should I change all unique_key examples on the snapshot_configs page to become arrays? IOW, is it understood that if you use an array you can also use a string? https://docs.getdbt.com/reference/snapshot-configs

[gshank]

1. Yes.
2. Yes, this replaces concatenating columns as unique keys
3. The inline notation should be fine too, it's just a matter of yaml syntax style.
4. Not backwards compatible. This is a net-new feature which is only going into 1.9 and versionless.
5. I don't think it's necessary to change all unique_key examples, but I would probably replace the concatenating column examples. Either a string or an array is acceptable. We do this in a number of places (that I can't think of right now...)