fix: improve instance wake/provision error quality

[bmdavis419]

Summary

• add step-aware context for wake/provision failures
• add server start retry behavior and richer diagnostics
• improve failure messaging with health/log-tail context

Issues

• Closes [Error]: Failed to warm up instance #164

Greptile Summary

This PR improves error diagnostics and reliability for instance wake/provision operations. Key changes include adding retry logic for btca server starts (up to 2 attempts with process cleanup between retries), optional git authentication configuration via BTCA_GIT_TOKEN env var, and enhanced error messages that include health check attempts/status and log tail output for easier debugging.

Key Changes:

• Added startBtcaServerWithRetry() function that retries server start up to 2 times with pkill cleanup between attempts
• Added configureSandboxGitAuth() to configure git credentials in sandbox (optional, requires BTCA_GIT_TOKEN env var)
• Enhanced error messages in provision/wake flows to include health check details and btca server log tail
• Added configure_git_auth step label for better error context
• Applied retry logic consistently across provision, wake, update, and createSandboxFromScratch flows

No breaking changes detected. All new functionality is additive and properly wrapped in error handling.

Confidence Score: 4/5

• This PR is safe to merge with minimal risk
• The changes improve error handling and add retry logic without breaking existing functionality. All new features are optional and well-integrated into existing error handling patterns. Minor deduction for lack of explicit validation on the git token format.
• No files require special attention

Important Files Changed

Filename	Overview
apps/web/src/convex/instances/actions.ts	Added git auth configuration, server start retry logic with pkill cleanup, and enhanced error messages with health check details and log tail context

_{Last reviewed commit: aec0baf}