feat: Add Chap-compatible CSV export with global temporal continuity [CLIM-360]

[abyot]

This PR adds a Chap-compatible CSV export utility for producing training datasets that can be consumed directly by Chap.

What’s included

• A CSV exporter that maps harmonized DataFrames to the Chap schema:

  • required and also reserved fields: time_period, location, disease_cases
  • optional reserved field: population
  • all other columns are treated as covariates

• Enforcement of global temporal continuity across all locations:

  • all locations must cover the same start/end time window
  • the window is inferred from the data or can be provided explicitly
  • missing periods are never imputed

• A diagnostic helper (find_temporal_gaps) that returns the complete list of missing periods per location to support upstream data repair or imputation

• Time-period normalization and validation for monthly and weekly data

• Configurable continuity handling at export time (error, warn, ignore)

Design notes

The exporter is intentionally strict and opinionated: its goal is to produce CSVs that are immediately usable by Chap without relying on implicit padding, masking, or downstream fixes.

Data completeness and imputation decisions are expected to happen upstream and are supported via explicit diagnostics rather than silent behavior.

Tests

Unit tests cover schema enforcement, time-period normalization, global continuity semantics, continuity policies, and successful CSV/file output.

[karimbahgat]

Looks great and seems very robust, with plenty of tests. Added some commits to simplify type hints and minor tweaks for code clarity. Ready for merge.