bugfix: Fix Test VCD Output

[atluft]

Support VCD dumps for IO registers with write handlers (#486), implements changes suggested by @jamesrwaugh

Testing

Running these commands

make
cd simavr
./run_avr ../tests/atmega88_example.axf
gtkwave gtkwave_trace.vcd

Output

./run_avr ../tests/atmega88_example.axf 
Loaded 1722 bytes of Flash data at 0
Loaded 114 bytes of Flash data at 0x6ba
Read from eeprom 0xdeadbeef -- should be 0xdeadbeef..
Read from eeprom 0xcafef00d -- should be 0xcafef00d..

[gatk555]

This looks good in principle, but there are a couple of issues:

If _avr_set_r() is called for a register with a direct write callback set, the following IRQ call will pass the value written by the CPU, not the actual register value. (Important example: interrupt flag registers with write-1-to-clear behaviour.) If the IRQ is called with the final value, there is then no need to modify _call_register_irqs(). Will it still work for the VCD file?

test_atmega168_ioport.tst is failing. The fault may be in the recently-modified test code.

[atluft]

@gatk555, thank you for the helpful comments.

Have a look at the changes in 1d323bf , passing all tests and VCD working.
I removed a code comment without understanding, maybe you can advise on the consequences of this approach?

       // The byte to be sent should NOT be written there,
       // the value written could never be read back.
       // avr_core_watch_write(avr, addr, v);