feat(grader): add task-aware grading with new metrics and composite scoring

refactor(simulator): add feedback on action effects and enforce capacity discipline

feat: implement Kubernetes executor for automated cluster scaling and infrastructure management

upgraded to 10 nodes

feat(curriculum): add progressive training curriculum management

fixed task 1, 2, 3

feat: implement core SRE simulation environment, Pydantic schemas, and physics models for task-based cluster management

kinda almost fixed task 1

normalized, better grading and env

tmp

update cost model and observation descriptions for clarity and accuracy

integrate production simulator, stability math, and reward recalibration