Replaced the stale system stack switch pseudo-code in the ARMv7-A ports with comments that describe what the code actually does

[fdesbiens]

Problem

The context save, vectored context save and system return routines in the ARMv7-A ports carry pseudo-code comments claiming that they save the thread stack pointer and then switch to _tx_thread_system_stack_ptr:

@    /* Save the current stack pointer in the thread's control block.  */
@    _tx_thread_current_ptr[core] -> tx_thread_stack_ptr =  sp;
@
@    /* Switch to the system stack.  */
@    sp =  _tx_thread_system_stack_ptr;
@
    MOV     r10, #0                             @ Clear stack limit

Neither of those things happens. None of these ports references _tx_thread_system_stack_ptr anywhere outside an unused IMPORT/.global in their example builds. The comments are leftovers from an older generic port template.

They are misleading rather than merely redundant: in #240 a user set out to add the "missing" stack switch code that these comments describe. This PR replaces them with a description of the mechanism the ports actually use.

Why there is no software stack switch on ARMv7-A

On ARMv7-A every processor mode has its own banked stack pointer, so the switch is done by the hardware.

• IRQ_Handler branches straight to _tx_thread_context_save while still in IRQ mode (ports_smp/cortex_a7_smp/gnu/example_build/startup.S:170), so on entry sp is already this core's banked IRQ stack, carved out of _stack_init_irq by startup.S:284-289. That stack is the system stack.
• The thread's stack pointer is stored in the control block by _tx_thread_context_restore, on the __tx_thread_preempt_restore path only (tx_thread_context_restore.S:197-233), and therefore only when the interrupt results in preemption.
• The scheduler runs on the banked SVC mode stack that startup.S:298-301 sets up.

The AArch64 SMP ports are deliberately left alone. ARMv8-A does not bank a stack pointer per processor mode, so those ports do keep _tx_thread_system_stack_ptr[core] as a real array and reload it explicitly; there the comments are accurate.

Scope

38 files, all comment-only.

Tree	Files	How
ports_arch/ARMv7-A/threadx/common/src/tx_thread_system_return.S	1	edited directly
ports/cortex_a{5,7,8,9,12,15,17}/{ac6,gnu}/src/tx_thread_system_return.S	14	regenerated with ports_arch/ARMv7-A/update.sh --port-sets tx
ports_module/cortex_a7/{gnu,iar}/module_manager/src/tx_thread_system_return.*	2	edited directly, not covered by a generator
ports_smp/cortex_a{5,7,9}_smp/{ac5,gnu} and ports_smp/cortex_r8_smp/ac5, three routines each	21	edited directly; there is no ports_arch/ARMv7-A/threadx_smp tree

Untouched, because their comments match their code: every ARMv8-A, ARC, MIPS, RISC-V, Xtensa, ARM9 and ARM11 port, and ports/cortex_a{5,7,8,9}/{ac5,ghs,iar}.

Validation

• Every added and removed line in the diff is a comment. There is no instruction change anywhere in the PR.
• All 25 GNU variants assemble cleanly with arm-none-eabi-gcc for their target core. The ac5 and iar files have no toolchain available here and were reviewed by diff.
• scripts/check_ports.sh passes, including the "generated ports are reproducible from ports_arch" check, which confirms the 14 regenerated files match the template exactly.

Known leftover, deliberately out of scope

Seven ARMv7-A SMP example builds still declare _tx_thread_system_stack_ptr with an unused IMPORT or .global. The symbol is genuinely defined in common_smp/src/tx_thread_initialize.c:43, so these link fine; removing them would mean touching example build files for no benefit.

Fixes #734
Follows up on #240