Add CI/CD pipeline and fix semihosting exit

[jserv]

Build system:

• Split check target into Arm Cortex-M (clean-build + baked flags) and POSIX (env var override at runtime) variants with shared test loop
• Add stale-clone guards before git clone in RTOS and Kconfig recipes
• Handle root-owned build artifacts in clean-build and distclean via sudo fallback when unprivileged rm fails
• Bump Cortex-M CHECK_TIMEOUT from 30 to 120 for slow CI runners
• Make defconfigs fully explicit with all config options listed

Semihosting exit:

• Add tm_semihosting_exit() in tm_putchar.c: direct ARM SYS_EXIT (0x18) via bkpt, bypassing newlib _exit() which pulls in __sinit and file I/O that depend on initialise_monitor_handles (omitted by -nostartfiles)
• Replace _exit() calls with tm_semihosting_exit() in tm_report.c
• Guard <unistd.h> include behind #ifndef TM_SEMIHOSTING

QEMU runner (scripts/qemu-run.sh):

• Replace timeout(1)/gtimeout with shell-managed watchdog to avoid false timeout reports from wrappers that mis-handle semihosting exit
• Redirect watchdog subshell stdout/stderr to /dev/null so $() capture in the Makefile check loop does not block for the full sleep duration
• Add cleanup trap (EXIT/INT/TERM) with proper signal handler split to prevent double-invocation and orphaned child processes
• Use mktemp sentinel file for timeout detection instead of signal codes, avoiding PID-reuse and external-kill misclassification

CI (.github/workflows/main.yml):

• coding-style: clang-format-20 for C, shfmt for shell scripts, trailing-newline check; PR-mode diffs only changed files
• posix-tests: 2x2 matrix (ThreadX/FreeRTOS x Ubuntu/macOS)
• cortex-m-tests: ThreadX and FreeRTOS under QEMU mps2-an385
• sanitizers: ASan+UBSan on POSIX builds (runs after posix-tests)