fix(runners, server): prevent CLI error traceback loss and missing failure details

[AkshatSoni26]

Problem & Motivation

1. Loss of Error Tracebacks: In CLI subprocess executions, critical error details (Python tracebacks, assertion failures, exit summaries) almost always appear at the end of stderr/stdout. The previous head-slicing approach ([:2000]) drops the actual root-cause traceback whenever log output exceeds 2,000 characters.
2. Loss of Failure Context: In runners.py, non-zero exit codes with stdout present were bypassing the command_failed error envelope, and in server.py, useful stdout execution hints were omitted from command failure details.

---

Solution & Changes

• Hybrid Log Truncation (truncate_output):
  • Preserves initial command invocation context (first 1,000 chars).
  • Adds an explicit truncation marker (... [N chars truncated] ...).
  • Preserves trailing stack trace / error root cause (last 1,000 chars).
• Consolidated Error Envelopes (run_json & _wrap_cli):
  • Non-zero exit status (code != 0) consistently returns command_failed.
  • Both stdout and stderr are safely truncated using truncate_output() and attached to error.detail.
• Unit Tests:
  • Added test coverage in test_runners.py for truncate_output (empty strings, under-limit, over-limit, custom limits, long output).
  • Added test coverage in test_server.py verifying _wrap_cli preserves stdout details on command failure.

---

Prior Art & Credit

This PR builds on the motivation explored in open PR #5 and PR #6 by @abhinavgautam01 (preserving stdout on command failures), extending the pattern by centralizing log truncation into a reusable, tested helper function.