[Code scan] Preserve installer failures through toolchain wrapper logging pipelines

[njzjz]

This issue is a result of a Codex global repository scan.

Several toolchain wrapper scripts pipe the installer into tee compile.log without enabling pipefail or returning the installer's ${PIPESTATUS[0]}. In Bash, the pipeline exit status is the status of tee, so CI/users can see a successful wrapper exit even when install_abacus_toolchain_new.sh failed.

Example GNU wrapper:

abacus-develop/toolchain/toolchain_gnu.sh

Lines 118 to 123 in 84ca04b

	${DRY_RUN_MODE:+$([ "$DRY_RUN_MODE" = "yes" ] && echo "--dry-run")} \
	${PACK_RUN_MODE:+$([ "$PACK_RUN_MODE" = "yes" ] && echo "--pack-run")} \
	${ENABLE_CUDA:+--enable-cuda} \
	${GPU_VERSION:+--gpu-ver="$GPU_VERSION"} \
	"$@" \
	| tee compile.log

Example Intel wrapper:

abacus-develop/toolchain/toolchain_intel.sh

Lines 135 to 140 in 84ca04b

	${DRY_RUN_MODE:+$([ "$DRY_RUN_MODE" = "yes" ] && echo "--dry-run")} \
	${PACK_RUN_MODE:+$([ "$PACK_RUN_MODE" = "yes" ] && echo "--pack-run")} \
	${ENABLE_CUDA:+--enable-cuda} \
	${GPU_VERSION:+--gpu-ver="$GPU_VERSION"} \
	"$@" \
	| tee compile.log

The same pattern exists in the GCC/MKL, GCC/AOCL, and AOCC/AOCL wrappers.

Relevant code:

"$@" \
  | tee compile.log

Suggested fix:

Enable set -o pipefail before the pipeline, or explicitly exit ${PIPESTATUS[0]} after tee so wrapper callers receive the installer failure status.