feat: improve browsecomp

[yuki-97]

Summary

Improves the browsecomp_agent with better observability and tighter control over long agentic trajectories.

• max_reset_count: Adds a new config field to cap the number of context resets per trajectory. Previously, context resets could happen unboundedly whenever the prompt exceeded the reset threshold. This lets users trade off context freshness against reset overhead.
• reset_count / num_tool_calls logging: Surfaces reset_count and num_tool_calls directly on the agent response object for downstream analysis. Also fixes num_tool_calls — it was previously counted inside the resources server and is now correctly tracked in the agent loop.
• Trajectory recording (snap_dir): Adds an optional snap_dir config field. When set, the agent saves a JSONL snapshot of the full conversation at each context reset and at the end of the trajectory (keyed by task_index and attempt). Useful for debugging and offline analysis of long browsing runs.
• Config: Exposes snap_dir and max_reset_count in benchmarks/browsecomp/config.yaml.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.