feat: implement max_response_time_ms behavior rule

[spboyer]

Summary

Implements max_response_time_ms behavior rule as requested in #136.

Working as Linus (Backend Developer).

Changes

internal/models/testcase.go

• Added MaxResponseTimeMs int64 field to BehaviorRules struct

internal/metrics/behavior.go

• Added MaxResponseTimeMs, ActualResponseTimeMs, MaxResponseTimeMsPassed fields to BehaviorMetrics
• Implemented timing compliance check in ComputeBehaviorMetrics: compares run.DurationMs against rules.MaxResponseTimeMs when set (>0)
• Included MaxResponseTimeMsPassed in AllConstraintsPassed()
• Updated computeEfficiency from 4 categories × 0.25 to 5 categories × 0.20

internal/metrics/behavior_test.go

• Added 4 new test cases: under limit, at limit, exceeds limit, combined failure
• Updated all existing test expected efficiency scores for new 5-category weighting
• Added wantResponseTimePassed assertion to test struct

schemas/task.schema.json

• Added max_response_time_ms (integer, minimum: 1) to behaviorRules definition

Usage

expected:
  behavior:
    max_response_time_ms: 5000   # must complete within 5 seconds
    max_tool_calls: 10

Breaking Change Note

Efficiency scoring changed from 4×0.25 to 5×0.20 per constraint category. Anyone relying on exact efficiency score values will see different numbers (e.g., 1 failed constraint: 0.75 → 0.80).

Closes #136

[Copilot]

Pull request overview

Implements the max_response_time_ms behavior constraint so runs can be evaluated against a maximum allowed duration, and surfaces the results in behavior metrics and scoring.

Changes:

• Add max_response_time_ms to BehaviorRules and to the task JSON schema.
• Extend BehaviorMetrics to record actual duration, pass/fail for the response-time constraint, and include it in AllConstraintsPassed().
• Update behavior efficiency scoring to 5 constraint categories and expand unit tests accordingly.

Show a summary per file

File	Description
schemas/task.schema.json	Adds max_response_time_ms to the behaviorRules schema.
internal/models/testcase.go	Adds MaxResponseTimeMs to BehaviorRules (and also introduces an output_contains_any field in TestExpectation).
internal/metrics/behavior.go	Implements response-time compliance check and updates efficiency scoring weights.
internal/metrics/behavior_test.go	Adds response-time test cases and updates expected efficiency scores for 5-category weighting.

Copilot's findings

• Files reviewed: 4/4 changed files
• Comments generated: 1

[spboyer]

Re: Copilot review on MayInclude/output_contains_any: This field is pre-existing and was not introduced by this PR. Our change only adds MaxResponseTimeMs to BehaviorRules. The schema alignment for output_contains_any can be tracked separately.

[Copilot]

Copilot's findings

• Files reviewed: 6/6 changed files
• Comments generated: 6