[CT-3078] Improve diff displayed in stdout when a test fails

[MichelleArk]

Description

Currently, a failing unit test simply logs both the actual and expected results to stdout:

❯ dbt unit-test
20:34:18  Running with dbt=1.7.0-b1
20:34:18  target not specified in profile 'postgres', using 'default'
20:34:18  Registered adapter: postgres=1.7.0-b1
20:34:19  Found 10 models, 3 seeds, 20 tests, 0 sources, 0 exposures, 0 metrics, 360 macros, 0 groups, 0 semantic models, 5 unit tests
20:34:19  Found 1 unit test, 2 models, 0 sources, 0 exposures, 0 metrics, 360 macros, 0 groups, 0 semantic models
20:34:19  
20:34:19  Concurrency: 1 threads (target='default')
20:34:19  
20:34:19  1 of 1 START unit_test my_model__test_my_model ................................. [RUN]
20:34:19  1 of 1 FAIL 1 my_model__test_my_model .......................................... [FAIL 1 in 0.11s]
20:34:19  
20:34:19  Finished running 1 unit_test in 0 hours 0 minutes and 0.53 seconds (0.53s).
20:34:19  
20:34:19  Completed with 1 error and 0 warnings:
20:34:19  
20:34:19  Failure in unit_test my_model__test_my_model (models/unit_testing/schema.yml)
20:34:19    

Actual:
| c |
| - |
| 3 |

Expected:
| c |
| - |
| 4 |

20:34:19  
20:34:19    compiled code at target/compiled/jaffle_shop/models/unit_testing/schema.yml/models/unit_testing/my_model__test_my_model.sql
20:34:19  
20:34:19  Done. PASS=0 WARN=0 ERROR=1 SKIP=0 TOTAL=1

This may work fine for very small expected results (few rows & columns), but would be difficult to debug for larger results as the diff is not visualized. For example:

❯ dbt unit-test
...
Found 1 differing row between actual and expected outputs for my_model:
Actual
- {'c': 3}
          ^
Expected
+ {'c': 4}

Acceptance Criteria

Test failure message displays diff in a readable format

Impact to other teams

None

Will backports be required?

No

Context

[dbeatty10]

Of the formats I've seen thus far, the daff format is the one I prefer since it makes it easier to see exactly where the differences are. The tabular diff specification for daff is described in more detail here.

[graciegoheen]

Since we're allowing folks to define fixtures in separate csv files, I think it would be good to include the input mock data in the logs as well so they don't have to go splunking through a ton of files.

[MichelleArk]

@dbeatty10 - I had a chance to spike how this might work with daff here: #8984!

[dbeatty10]

@dbeatty10 - I had a chance to spike how this might work with daff here: #8984!

Ooooh... 🤩 beautiful output and elegant implementation!

[graciegoheen]

From @dbeatty10 for docs

We should let consumers know that the output format is an experimental and unstable interface designed for human eyes (rather than machines) and that we reserve the right to change the diff format that is printed to the terminal and in the logs.