UKSingleYearDataset silently ignores fiscal_year with file_path, and defaults it to a hardcoded 2025

UKSingleYearDataset.__init__ (policyengine_uk/data/dataset_schema.py:42-67) takes fiscal_year: int = 2025, but the argument is only used on the DataFrame branch:

if file_path is not None:
    ...
    self.time_period = str(f["time_period"].iloc[0])   # line 58
else:
    ...
    self.time_period = str(fiscal_year)                # line 67

Two problems.

fiscal_year is silently ignored when file_path is given. Passing a conflicting value has no effect and produces no warning:

frs_2024_25.h5                        -> time_period = 2024
  same file with fiscal_year=2030     -> time_period = 2024   (argument ignored)

The default is a hardcoded year that drifts. Building from DataFrames without passing fiscal_year silently labels the data 2025:

UKSingleYearDataset(person=..., benunit=..., household=...)  -> time_period = 2025

The latest FRS release is 2024/25, which on our convention is label 2024 (CURRENT_FRS_RELEASE.survey_year = 2024), so the default is already a year ahead of the newest data available, and it will be wrong differently every year. An external collaborator raised exactly this.

Nothing we ship is affected — create_frs passes fiscal_year=year at all three construction sites — but the failure mode for anyone assembling a dataset by hand is a silently mislabelled time_period, which then feeds the wrong policy year into every calculation.

Suggested fix: warn when fiscal_year is passed alongside file_path (where it cannot take effect), and make fiscal_year required on the DataFrame branch rather than defaulting to a hardcoded year.