Contrib data handling

[bfoley12]

Summary

Reworks how Contribution.data is parsed, validated, and stored, and implements throughout CRUD paths. Also introduces pivoting a single contribution into multiple based on the presence of conditions in the data field, SI unit canonicalization, bulk-updates of Contributions (currently only used to efficiently mark many contributions as public), and string-normalization types.

Major changes:

Domain changes

• domains/contributions/data.py: Contribution.data parses annotated keys of the form: "col_name (unit, cond1=a, cond2=b)" into properly formatted data entries.
  • "properly formatted" means we store:
    • original unit, value, error,
    • the SI converted unit, value, and error
    • precision (sig figs)
• domains/contributions/pivot.py: expands one contribution whose data keys carry conditions into multiple contributions, keyed by condition_key
• domains/contributions/stats.py: used to derive Project.stats and Project.columns from contributions
• domains/_shared/units.py: SI unit normalization (Pint) and error support (uncertainties)

Model changes

• Contribution: added condition_key as an identifier, removed needs_build, server generates _id
• Structure.Lattice.matrix: switched from PolarsFrame to LatticeMatrix (3x3 matrix) validation to skip polars-frame encoding
• Project: stats and columns are now derived by the server from Contributions rather than supplied, is_approved is now admin-only
• MPContribSettings: defining controls on model relations (max contribution.data depth, max_columns, max_components, float_precision)