Backcheck date column joins backcheck data on the survey KEY

[iabaako]

Stack position

• Order: 6 of 13
• Branch: fix/300-backcheck-date-join, from fix/299-backcheck-settings-fields
• PR base: fix/299-backcheck-settings-fields (retarget to main once Backcheck duplicate-handling option and target % are silently ignored #299 merges)
• Blocked by: Backcheck duplicate-handling option and target % are silently ignored #299
• Next: Backchecks: track progress against the backcheck target #318 branches from this branch

Description

_add_date_columns
(compute.py:747-751)
takes the survey_key column from the backcheck data and joins it against
the survey KEY in the analysis:

bc_dates = backcheck_data.select(
    [survey_key, pl.col(backcheck_date).alias("backcheck_date_col")]
).unique(subset=[survey_key])
result = result.join(bc_dates, on=survey_key, how="left")

Survey and backcheck submissions normally have different KEYs, so
backcheck_date_col comes back empty. It feeds the average-days figures in the
enumerator and backchecker statistics (_calculate_average_days).

Steps to reproduce

Use a project whose survey and backcheck datasets have distinct KEY values and a
configured backcheck date. The backcheck dates and the average-days statistics
are empty.

Fix

Join on the backcheck KEY ({survey_key}__BCCL) when it is present, as the
staff join does at
compute.py:703, or on
survey_id.

Acceptance criteria

• Backcheck dates populate when survey and backcheck KEYs differ.
• Tests use distinct survey and backcheck KEYs.