GM ignition: remove signal with noise and false positives

[sshane]

Linked to issue commaai/openpilot#23645

Verify on these cars:

• HOLDEN ASTRA RS-V BK 2017
• CHEVROLET MALIBU PREMIER 2017
• CADILLAC ATS Premium Performance 2018
• BUICK REGAL ESSENCE 2018 - no active users
• GMC ACADIA DENALI 2018 - there's only two active users, script checked one user and I manually verified the other using available recent rlogs (e38a02cae1ef1e0c|2022-01-31--18-21-53, e38a02cae1ef1e0c|2021-11-28--01-36-02, e38a02cae1ef1e0c|2021-11-27--13-35-57)

Thanks to a route provided by @twilsonco (e69bcbc2c84de4d2|2022-02-02--17-03-51), this shows the preconditioning feature of a Chevy Volt being used, and has the old signal at 1 and the new signal at 0. This is the reason there's so many mismatches in the Chevy Volt comparisons, they're all false positives. Plot: https://imgur.com/a/Ed2MHWk

A few plots: https://imgur.com/a/PEYtWm5

Checked data:

Platform: GMC ACADIA DENALI 2018
Checked 1 unique dongles, 53 unique routes, and 593 segments
Mismatches: 23 frames, 0 segments (any frame)
---
Platform: CADILLAC ESCALADE ESV 2016
Checked 10 unique dongles, 116 unique routes, and 617 segments
Mismatches: 0 frames, 0 segments (any frame)
---
Platform: CHEVROLET VOLT PREMIER 2017
Checked 45 unique dongles, 1464 unique routes, and 14022 segments
Mismatches: 751821 frames, 231 segments (any frame)

[adeebshihadeh]

I checked off the first three, since we moved them to dashcam only recently until they have users again. https://github.com/commaai/openpilot/blob/a39e635b5ce271738504abaa84d060c7b26206c0/selfdrive/car/gm/interface.py#L50