FIX - Shm truncation guard

[Sahil-u07]

Summary

Fix a correctness bug in the C++ shared-memory write path that previously logged a warning and continued writing truncated payloads. Oversized payloads now fail fast by throwing a std::runtime_error, preventing silently corrupted data from flowing through a running study.

Changes

• concore.hpp
  • Added an explanatory comment above SHM_SIZE.
  • Replaced warning-and-truncate behavior with a std::runtime_error when the payload size is >= SHM_SIZE in both shared-memory write paths.
• test_shm_abort.py
  • New focused regression tests that compile a minimal C++ driver and verify oversized payloads fail while in-range payloads succeed.

Why this matters

Silent truncation of SHM payloads can corrupt downstream neuromodulation calculations and produce incorrect scientific results without any visible error in running studies. This change forces a visible failure so users notice and fix configuration (increase SHM_SIZE or reduce payloads) rather than producing silent, incorrect outputs.

Testing / CI

• Ensure CI runs the new tests on a Linux runner with g++ installed.
• Expected results on Linux:
  • test_oversized_payload_throws — non-zero exit, stderr contains "Aborting", no "truncated".
  • test_within_limit_succeeds — exit 0, no stderr.
  • test_exactly_at_limit_throws — non-zero exit.
  • test_one_under_limit_succeeds — exit 0, no stderr.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.