feat: populate Rows for SS04 and SS10 validation errors

[javihern98]

Summary

• SS04 (codelist mismatch) and SS10 (cube constraint) errors now populate the Rows field with the dataframe records holding each offending value, matching the behaviour already provided by SS02/SS05/SS06/SS07/SS11.
• create_error_SS10_SS04 gains a leading data (DataFrame) parameter; both call sites in process_errors_by_column are updated. Public API change — external callers passing positional args need to insert data before values.
• Bumps version 2.6.8 → 2.6.9.

Motivation

Until now, when a CSV contained a wrong codelist value (e.g. INV_ADV_TY=TNLSCOD where TNLSCOD is not in the codelist), the resulting SS04 warning told you what the bad value was but not which rows contained it. Downstream consumers had to re-scan the dataframe themselves with brittle message parsing.

After this change, the warning carries the same shape as SS07:

{
  "Code": "SS04",
  "Component": "INV_ADV_TY",
  "Rows": [{ "FREQ": "M", "INV_ADV_TY": "TNLSCOD", "OBS_VALUE": "230", ... }, ...],
  "Message": "Wrong value TNLSCOD for dimension INV_ADV_TY"
}

Verified on a downstream consumer: one bad value (TNLSCOD) yields one SS04 warning whose Rows lists all 80 dataframe records where INV_ADV_TY == "TNLSCOD". Per-record dict shape is identical to what SS07 produces.

Files changed

• sdmxthon/parsers/data_validations.py — function signature + body, two call sites
• sdmxthon/__version__.py — 2.6.8 → 2.6.9
• Changelog.rst — new entry under 2.6.9

Test plan

• Existing unit tests pass (no in-repo callers besides the two updated ones)
• Validate a CSV that triggers SS04 — confirm Rows is a list of dicts, one entry per matching dataframe row
• Validate a CSV that triggers SS10 — same confirmation
• Confirm SS02/SS05/SS06/SS07/SS11 still populate Rows correctly (no regression in the unmodified code paths)