[Volkopat]

Fixes #24605.

ZEND_CHECK_STACK_LIMIT guards the recursion in serialize(), var_dump() and debug_zval_dump(),
but not in var_export(), print_r() or unserialize(). On a structure that fits the stack those
functions recurse until it is exhausted and the process dies with SIGSEGV - no error, no message.

Reproduced first

Each function run in its own process, stock php:8.5-cli, --ulimit stack=524288:524288, 3,000 levels:

call	result before the fix
serialize()	throws Error (already guarded)
var_export()	exit 139, SIGSEGV
print_r()	exit 139, SIGSEGV
unserialize()	exit 139, SIGSEGV

Change

The same guard the already-fixed paths use, at each recursive entry point:

• ext/standard/var.c - php_var_export_ex() (return FAILURE;)
• Zend/zend.c - zend_print_zval_r_to_buf(), the recursion behind print_r(), which lives in Zend
  rather than ext/standard (return;)
• ext/standard/var_unserializer.re - php_var_unserialize_internal() (return 0;)

Each site returns the type its function already uses, so no signatures change. 3 files, +18 lines.

#ifdef ZEND_CHECK_STACK_LIMIT
	if (UNEXPECTED(zend_call_stack_overflowed(EG(stack_limit)))) {
		zend_call_stack_size_error();
		return FAILURE;
	}
#endif

After

Built from current master with the patch applied, same structure, same 512 KB stack:

--- serialize (already guarded) ---
  serialize threw: Maximum call stack size of 475136 bytes (zend.max_allowed_stack_size - zend.reserved_stack_size) reached. Infinite recursion?
--- var_export ---
  var_export threw: Maximum call stack size of 475136 bytes ... reached. Infinite recursion?
--- print_r ---
  print_r threw: Maximum call stack size of 475136 bytes ... reached. Infinite recursion?
--- unserialize ---
  unserialize threw: Maximum call stack size of 475136 bytes ... reached. Infinite recursion?

All four now raise the same catchable Error, and the surrounding behaviour is unchanged for
structures that fit.

Scope I did not test

unserialize()'s own max_depth option (default 4096) usually rejects deep input first, so the new
guard only matters when the stack is small relative to that limit - which is the case tested here.
I did not test a very large max_depth against a large stack.