Add ATmega8 and fix some minor bugs with program-avr-spi applet

[turmoni]

There are two changes here:

1. Adding the ATmega8 to the database of AVR microcontrollers. erase_time was chosen to match the other entry with an erase_time, since not having one seemed to lead to issues with verifying what had been written, and this one seems to work fine
2. Fixing up the program-avr-spi applet so it doesn't give a traceback whenever you're doing something with an unidentified device

For an explanation of the second one, see the following output I got before adding the ATmega8:

% glasgow run program-avr-spi -V 5 --pin-reset 0 --pin-sck 1 --pin-cipo 2 --pin-copi 3 identify
I: g.device.hardware: device already has bitstream ID 57b5cdebb6d0d20034246a56c317b332
I: g.cli: running handler for applet 'program-avr-spi'
I: g.applet.program.avr.spi: port(s) A, B voltage set to 5.0 V
I: g.applet.program.avr.spi: device signature: 1e 93 07 (unknown)
Traceback (most recent call last):
  File "/home/dave/.local/bin/glasgow", line 8, in <module>
    sys.exit(run_main())
             ~~~~~~~~^^
  File "/home/dave/git/glasgow/software/glasgow/cli.py", line 942, in run_main
    exit(asyncio.new_event_loop().run_until_complete(main()))
         ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^
  File "/usr/lib/python3.13/asyncio/base_events.py", line 720, in run_until_complete
    return future.result()
           ~~~~~~~~~~~~~^^
  File "/home/dave/git/glasgow/software/glasgow/cli.py", line 710, in main
    return applet_task.result()
           ~~~~~~~~~~~~~~~~~~^^
  File "/home/dave/git/glasgow/software/glasgow/cli.py", line 659, in run_applet
    return await applet.interact(device, args, iface)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dave/git/glasgow/software/glasgow/applet/program/avr/__init__.py", line 259, in interact
    if device.erase_time is not None:
       ^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'erase_time'

This is actually the result of two different things that needed to be fixed:

1. The code was attempting to use the device object before it checked to see if the device existed (so this change brings that further up)
2. The code that checks for the existence of the device explicitly excludes checking when either called with no action or with the "identify" action, which means that you would still get this error even with 1 being fixed

I don't know if a simple return is what's wanted here, but I assume the point of the check was to not error out on an unidentified device because it's already done everything that's needed for the identify command

[whitequark]

This skips over iface.programming_disable().

[whitequark]

I'd say add programming_disable after the identify call here, and enable it once we operate on a known device.

[turmoni]

Hrm, I'm wondering if it might be worth explicitly calling programming_disable before the return, and making ProgramAVRError also call programming_disable, unless I'm misinterpreting the reason for this suggestion or how the code works (both are entirely possible). I think any exception will leave it in this same state.

[whitequark]

unless I'm misinterpreting the reason for this suggestion or how the code works (both are entirely possible)

There's a programming_disable call at the very end that your return was skipping over.

[turmoni]

Yeah, but I think that any raised ProgramAVRError will also miss the programming_disable call, unless I'm misunderstanding. If it is doing that, would it be worth calling programming_disable on all of the errors and explicitly doing it early for the case where the operation is blank or identify?

[whitequark]

How about putting the body of the function into a try..finally block which will call programming_disable()? You'd want the initialization of the interface and the initial programming_enable() command to be outside of the block, of course.