Study 09 author baseline: Beron (mouse) RFLR

## Context

Parent: #134. Beron (mouse) is a primary Study 09 cohort, but Result 1 currently lacks the paper's recursively formulated logistic-regression (RFLR) baseline.

## Done when

- [ ] Pin the RFLR equations, associated code revision, parameter bounds, condition handling, and immutable references.
- [ ] Implement the RFLR likelihood in `aind-dynamic-foraging-models` with equation-level and session-reset tests.
- [ ] Fit one model per subject on the existing adaptation sessions and score the identical immutable test sessions.
- [ ] Verify exact held-out trial-key parity with common Q and GRU.
- [ ] Freeze run/artifact provenance and add Beron (mouse) to Study 09 Result 1.
- [ ] Explain the RFLR/RL/sticky-HMM equivalence boundary and report reproduction confidence.

## Notes

Use the paper's RFLR parameterization as the named author baseline; do not substitute a mathematically related model without documenting it.