Add edge_pipeline_config parameter to detector creation

[timmarkhuff]

Summary

Adds support for the edge_pipeline_config parameter when creating detectors, mirroring the existing pipeline_config parameter. This allows callers to specify which pipeline config to use for edge inference instead of always using the mode's default.

Corresponds to the backend change in zuuul, PR #6056.

Changes

• Updated spec/public-api.yaml with the new edge_pipeline_config field on DetectorCreationInputRequest
• Ran make generate to regenerate the OpenAPI client models
• Added edge_pipeline_config parameter to all detector creation methods in client.py and experimental_api.py:
  • _prep_create_detector
  • create_detector
  • get_or_create_detector
  • create_binary_detector
  • create_counting_detector
  • create_multiclass_detector
  • create_bounding_box_detector
  • create_text_recognition_detector (ExperimentalApi)

Testing

Tested end-to-end against an eksdev environment running the zuuul branch from #6056:

from groundlight import Groundlight

gl = Groundlight()

det = gl.create_detector(
    name="edge-config-sdk-test",
    query="Is this image relevant?",
    pipeline_config="never-review",
    edge_pipeline_config="generic-cached-timm-efficientnetv2s-knn",
)
print(f"Created detector: {det.id}")
# Created detector: det_3AuicidzisgV5ja7fFhenB8RSIL

[brandon-wada]

LGTM