OD calibration sharing and pooling feature

[fabiansisteffen]

OD calibration sharing and pooling feature

The cluster-wide OD calibration sharing and pooling feature with 2 operations or large and small cluster scenarios.

• copy calibration
• pool calibration

core/pioreactor/calibrations/pooling.py (New)

• Added pool_od_calibrations, which validates that incoming OD calibrations share the same angle, pd_channel, and a highly similar ir_led_intensity (within 5%).
• Points from all compatible donors' recorded_data are merged into a single array and re-fit using the existing spline_fit (or other supported methods).
• Designed for extensibility via the _POOLING_HANDLERS registry so pump/stirring can easily be added later.

core/pioreactor/web/unit_api.py (Worker API Updates)

• Added GET /calibrations//active which returns the raw YAML of a unit's active calibration.
• Added GET /calibrations//<calibration_name> to allow specific calibration retrieval by name.
• Updated POST /calibrations/ to enforce the "auto-activation guard": it only automatically activates the newly pushed calibration if the target unit didn't already have an active one for that device.

core/pioreactor/web/api.py (Leader API Updates)

• Added POST /cluster/calibrations//pool: Fans out to get active calibrations from workers, pools them, and returns the merged YAML.
• Added POST /workers/<pioreactor_unit>/calibrations//apply: Takes a calibration YAML payload and pushes it to a single unit or the whole cluster ($broadcast).
• Added POST /workers/<source_unit>/calibrations//copy_to/<target_unit>: A convenience method to directly copy an active (or specific named) calibration from one unit to another (or all units).

core/tests/test_calibration_pooling.py

• Test covering merging logic, mismatch rejection (wrong angle, different LED intensity), single-donor behaviour, and the handler registry.