support snapshot option 'invalidate_hard_deletes'

[willi-mueller]

This PR:

1. adds the option to invalidate_hard_deletes which marks records missing in the source as deleted in the snapshot
2. adapts snapshot_merge.sql so that it follows the same structure as the file in dbt-core

Tests

I tested snapshots manually both with and without the new configuration option invalidate_hard_deletes enabled.

{{
    config(
      unique_key='id',
      target_schema='test_schema',
      strategy='check',
      invalidate_hard_deletes=True,
      check_cols=[ 'val' ]
    )
}}

Step 0: One row (id = 1) is present in the source

Step 1: A second row (id = 2) is present in the source

Step 2: The first row (id = 1) is deleted in the source. Thus, valid_to is set by the dbt snapshot run

Step 3: The first row (id = 1) appears again in the source. Thus, it is added to the snapshot as a valid row

Step 4: The second row (id = 2) changes its value to val = 'v2'. Thus, the old record is is marked as valid_until und a new row for the current value is appended to the snapshot.

Thus it has been demonstrated that this PR satisfies the following cases:

1. The existing functionality of tracking changes of rows over time has not been changed
2. Hard deletes in the source are tracked as well if the option is enabled. This means that if a row does not appear in the source data, it will be marked as deleted in the snapshot
3. Once deleted, rows can re-appear in the source and their re-appearance will be tracked in the snapshot as well.

Background

This PR aims to follow the code from dbt-core as seen here and here

[willi-mueller]

The code for this PR is complete.

I just want to add more information about my test protocol so that reviewers can reproduce it. Apart from that, the PR is done.

[tglunde]

all good