Armv7m: Fix stm/ldm

[dop-amin]

• stm was using the wrong registers to write to the reg list
• stm/ldm did not set obj.address so the selftest failed because the pointer could not be indentified
• Add an example to use/not use the splitting on the same piece of code to actually trigger ldm+stm

[mkannwischer]

* Example: add_kyber_m7 (dry run only)...
  File "/home/runner/work/slothy/slothy/example.py", line 2978, in <module>
    main()
  File "/home/runner/work/slothy/slothy/example.py", line 2971, in main
    run_example(e, debug=args.debug, dry_run=args.dry_run,
  File "/home/runner/work/slothy/slothy/example.py", line 2967, in run_example
    ex.run(**kwargs)
  File "/home/runner/work/slothy/slothy/example.py", line 165, in run
    self.core(slothy, *self.extra_args)
  File "/home/runner/work/slothy/slothy/example.py", line 2330, in core
    slothy.fusion_loop("1", ssa=False)
  File "/home/runner/work/slothy/slothy/slothy/core/slothy.py", line 462, in fusion_loop
    SourceLine.apply_indentation(self._fusion_core(pre, body, post, logger, **kwargs), indentation) + \
                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/slothy/slothy/slothy/core/slothy.py", line 414, in _fusion_core
    dfg.apply_fusion_cbs()
  File "/home/runner/work/slothy/slothy/slothy/core/dataflow.py", line 579, in apply_fusion_cbs
    return self.apply_cbs(fusion_cb, logger, one_a_time=True)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/slothy/slothy/slothy/core/dataflow.py", line 519, in apply_cbs
    if cb(t):
       ^^^^^
  File "/home/runner/work/slothy/slothy/slothy/core/dataflow.py", line 578, in fusion_cb
    return t.inst.global_fusion_cb(t, log=logger.info)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/slothy/slothy/slothy/targets/arm_v7m/arch_v7m.py", line 2005, in core
    ptr = inst.args_in_out[0]
          ~~~~~~~~~~~~~~~~^^^

I'm not sure why this is failing now. Is something rewriting those arguments? args_in_out[0] does look correct to me....

[dop-amin]

In my PR I set the obj.addr correctly in order for the selftest to recognize the pointer.
This will trigger ldm being affected by fixup.
This will make inst.args_in_out empty.
This makes the fusion fail as inst.args_in_out is used there.