Add significant_figures and decimal_places grading parameters - #277

Requires #273.

Problem

Custom criteria had no way to grade a numeric response to a fixed number of significant
figures or decimal places — both for correctness and for the precision the response was
actually written to (e.g. rejecting 3.1 or 3.14159 when 3.14 is asked for). There was
also no fuzzy variant that derives a numerical tolerance from a sig-fig/decimal-place count
without checking written precision.

This PR merges in #273 (significant_figures/sig_figs and
significant_figures_tolerance/sig_figs_tol, still open/unmerged) and adds its decimal-place
counterpart, decimal_places/dp and decimal_places_tolerance/dp_tol, reusing the same
numeric-string helpers. The two features are combined in one branch here because
decimal_places builds directly on the sig_figs helpers/integration points, and #273's branch
had drifted behind main's bracket-matching work (#265/#274) and atol unit-factor fix (#272) —
merging it in first and resolving those conflicts made it possible to build decimal_places on a
consistent base rather than duplicating the shared code path across two divergent branches.

Changes

• app/utility/expression_utilities.py: significant-figures helpers (round_to_sig_figs,
  split_numeric_string, count_sig_figs, sig_figs_match,
  relative_tolerance_from_sig_figs) from Add support for significant figures (sig_figs) parameter #273, plus new decimal-place helpers
  (round_to_decimal_places, count_decimal_places, decimal_places_match,
  absolute_tolerance_from_decimal_places) built on top of them.
• app/evaluation.py: parses significant_figures/sig_figs,
  significant_figures_tolerance/sig_figs_tol, decimal_places/dp and
  decimal_places_tolerance/dp_tol; all four are mutually exclusive with each other and with
  atol/rtol; sig_figs_tol writes rtol, dp_tol writes atol.
• app/context/symbolic.py, app/context/physical_quantity.py: integrate sig_figs/dp at
  the point each context decides a direct response = answer numeric match (before the
  ordinary equality/tolerance logic, since a value that's numerically equal but written to the
  wrong precision must still fail).
• app/docs/user.md, app/docs/dev.md: user-facing parameter docs and developer notes for all
  four parameters, including the caveat that decimal_places with physical_quantity compares
  at the answer's SI scale (so is intended for same-unit responses), unlike the scale-invariant
  significant_figures.
• Tests added across app/tests/expression_utilities_test.py,
  app/tests/symbolic_evaluation_test.py, app/tests/physical_quantity_evaluation_test.py for
  every new helper and both parameters (core cases, aliases, mutual-exclusion, invalid-value
  validation, physical-quantity/unit interaction).

Checklist

• Tests added/updated
• Docs updated (app/docs.md) if user-facing behaviour changed

Closes #

🤖 Generated with Claude Code

https://claude.ai/code/session_01PndQqZ1ZR2QQ1sYH3eNybZ