Interpolate time-varying params in reaction propensity_fn

[jc-macdonald]

Summary

Companion to #195 (source-only transitions), and independently useful on
its own -- found while wiring diphtheria_outbreakvacc's cross-district
case-importation term, which needs a time-varying hazard (concentrated in
an initial window, much lower afterward).

_build_reaction_artifacts compiles each named transition's propensity
independently of _wrap_eval_fn_for_time_varying /
_wrap_pytree_eval_fn_for_time_varying -- the two wrappers that make a
rate like rate: "lambda_import[time, loc]" work for the deterministic
RHS, by interpolating the raw (time, loc) grid down to the current
timestep's (loc,) slice before the compiled code runs (the expression
itself is already rewritten to the time-stripped lambda_import[loc]
form upstream by _strip_time_axis_in_expr, so the compiled code expects
the reduced shape). Reaction propensity_fns never went through either
wrapper, so the same rate reached a reaction's compiled code as the raw,
un-interpolated full grid array -- a shape mismatch. Confirmed via a
failing call before this fix:

ValueError: cannot reshape array of size N into shape (...)

Changes

• Adds _wrap_propensity_fn_for_time_varying, mirroring the existing two
  wrappers' exact interpolation logic (same _interp_along_axis call,
  same linear-interpolation/constant-extrapolation behavior), applied to
  every compiled reaction in _build_reaction_artifacts.

Test plan

• New test_time_varying_rate_propensity_matches_deterministic,
  cross-checking the reaction's propensity against the deterministic
  eval_fn's own (already-correct) interpolated inflow at a
  fractional, off-grid t -- same correctness-oracle pattern as the
  existing test_reactions_reconstruct_deterministic_eval_fn.
• Full test suite: 457 passed
• mypy --strict on compile.py: clean
• ruff check: no new findings beyond this project's existing
  unrelated pre-existing noqa notes