feat: add keypoint detection ingestion template

[shujaatTracebloc]

Summary

• Adds a new templates/keypoint_detection/ template with main ingestion script, sample images, CSV labels file with JSON keypoint annotations, and README documentation
• Follows the same conventions as existing templates (object_detection, image_classification)
• Uses TaskCategory.KEYPOINT_DETECTION with JSON-based Annotation and Visibility columns

Closes #52

Test plan

• Verify template structure matches other templates
• Verify keypoint_detection.py runs without import errors
• Verify sample images are valid JPGs with visible keypoint markers
• Verify CSV JSON columns parse correctly
• Verify annotation format aligns with client keypoint_detection use case expectations

🤖 Generated with Claude Code

---

Note

Medium Risk
Adds a new ingestion pathway and updates validator selection for SEMANTIC_SEGMENTATION/KEYPOINT_DETECTION, which can change runtime validation behavior for these task types. Also updates semantic segmentation template defaults and CSV headers, which may break existing template-based workflows if users relied on the prior column name or intent.

Overview
Adds a new templates/keypoint_detection example (script, README, and sample CSV) for ingesting image-based keypoint detection datasets using JSON Annotation/Visibility fields and TaskCategory.KEYPOINT_DETECTION.

Aligns the semantic segmentation template with the expected CSV header (filename instead of data_id) and switches the example ingestion intent to Intent.TRAIN.

Extends validators_mapping.map_validators to return dedicated validator sets for SEMANTIC_SEGMENTATION (including mask PNG validation) and for KEYPOINT_DETECTION (image file type + resolution checks), rather than falling back to the default mapping.

^{Reviewed by Cursor Bugbot for commit 2e6fc19. Bugbot is set up for automated code reviews on this repo. Configure here.}