Alignment to dev

[leoglonz]

Issue Addressed

Introduces a shared BasePhysicsModel interface for HBV models, deduplicating repeated code and adding new opt-in Hbv_2 features, plus fixes for regressions the refactor introduced.

Description

• Added hydrodl2.models.base.BasePhysicsModel and shared helpers in core/calc/utils.py (e.g. trim_warmup); hbv.py, hbv_1_1p.py, hbv_2.py, hbv_2_hourly.py now subclass it, removing large amounts of duplicated boilerplate across the four models.
• hbv_2.py: added opt-in flags elev_parTT (elevation-based parTT override), gage_agg (in-loop gage-level aggregation), all_output, and an optional routing_state carryover input.
• hbv_2_hourly.py: same BasePhysicsModel/trim_warmup refactor; removed dead self.initialize branch.
• api/methods.py: load_model() now also tries a PascalCase name conversion (e.g. hbv_adj -> HbvAdj) before falling back to the first class found in the module.
• Fixed two regressions surfaced by the refactor, verified against hydrodl2_master and against dmg's test suite:
  • load_model()'s fallback could resolve to the wrong class (e.g. BasePhysicsModel itself, or an unrelated imported Hbv variant) once modules started importing BasePhysicsModel/other classes. Fixed by adding a case-insensitive name-match step and restricting the last-resort fallback to classes actually defined in the target module.
  • elev_threshold (replacing elev_parTT) default was False, silently dropping the original model's always-on elevation-based parTT override with no config anywhere setting the key. Reverted default to True.
• hbv_2.py/hbv_2_hourly.py: SLZ lateral-flow clamp floor raised from min=0.0 to min=self.nearzero — min=0.0 lets SLZ collapse to exactly 0 for some parameter draws, permanently zeroing capillary rise and, in one observed case (dmg's test config), producing an all-zero/dead streamflow output.
• Added regression tests to tests/test_model_structure.py: load_model() identity checks for all loadable HBV variants, and an explicit default-flag check for Hbv_2 (elev_parTT/gage_agg/all_output) so a future default flip fails immediately instead of surfacing only as a numeric drift.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Code cleanup/refactor
• Documentation update

Other (please specify):

Checklist

• Branch is up to date with master
• Updated tests or added new tests
• Tests pass locally (117 passed, 1 skipped in this repo; 166 passed, 10 skipped in dmg's suite against this branch)
• Updated documentation (if applicable)
• Code follows established style and conventions