Add comps training data to `model.training_data`

[jeancochrane]

The model.training_data table is currently restricted to only final model runs:

data-architecture/dbt/models/model/model.training_data.py

Lines 14 to 23 in 236e8cc

	# Build the base metadata DataFrame
	base_query = """
	SELECT
	run_id,
	year,
	assessment_year,
	dvc_md5_training_data
	FROM model.metadata
	WHERE run_type = 'final'
	"""

However, it would also be useful to have it for comp runs, which would greatly reduce the maintenance burden of this CTE in the pinval.vw_comp view:

data-architecture/dbt/models/pinval/pinval.vw_comp.sql

Lines 39 to 55 in 236e8cc

	training_data AS (
	SELECT
	train.*,
	meta.model_predictor_all_name
	FROM {{ ref('model.training_data') }} AS train
	LEFT JOIN {{ source('model', 'metadata') }} AS meta
	ON train.run_id = meta.run_id
	-- Currently the `model.training_data` table only includes training data
	-- for final model runs, not comp runs, so we can only use final model runs
	-- here. Further, we have to make a manual decision about which final model
	-- run has training data that matches the comp run for assessment years
	-- that have multiple final models.
	WHERE train.run_id IN (
	'2024-03-17-stupefied-maya',
	'2025-02-11-charming-eric'
	)
	),

Tasks include:

• Add run_type = 'comps' to the conditional in the model.training_data query
• Rerun model.training_data locally and confirm that it adds rows for the two comps runs (2024 and 2025)
• Adjust the docs for model.training_data to clarify that comps runs should also be included
• Refactor pinval.vw_comp view to remove the WHERE clause in this CTE and adjust the join to join by run_id instead of assessment_year

Adapted from #864 (comment).