Use a more robust 6522 detection method

[tomcw]

Detection for each 6522: perhaps repeat 256 times, and internally record results as:

• 256 passes: 100% working
• 256 failures: not present
• otherwise: flakey

ie. if it's flakey, then visually show this detected state in the UI (eg. FLASHing, or suffix with "*" ?)

[tomcw]

For @roughana's card, he report:

1.54 started playing first time from menu, then not the 2nd time (quick exit), then played the 3rd time.

And from screenshot, sometimes the 6522-A can have a problem with data line 2 (0x04), sometimes with data line 3 (0x08) and other times it'll be fine.

EG. 1) Failed for "TEST: 50:04:00", this is showing that data line 2 reads or writes erroneously: Expected vs Actual:

• 55 vs 51 = 0101_0101 vs 0101_0001

2) Failed for "TEST: 50:05:00"... the expected vs actual values show that sometimes data line 3 reads erroneously as HIGH (or maybe writes with data line 3 HIGH?): Expected vs Actual:

• B6 vs BE = 10110110 vs 10111110
• 44 vs 4C = 01000100 vs 01001100

[tomcw]

EG, see this comment where I've got AppleWin progressing further (or failing sooner): https://github.com/tomcw/mb-audit/issues/25#issuecomment-2453556582