feat(cardputer-adv): add CC1101 nRF24 LoRa 3-in-1 expansion support

[leandrofleury]

Proposed Changes

Add support for the JosephCGS CC1101/NRF24/LoRa 3-in-1 expansion board on the M5Stack Cardputer ADV.

This introduces a dedicated PlatformIO environment:

m5stack-cardputer-adv-3in1

The standard m5stack-cardputer environment remains unchanged.

The expansion shares the Cardputer ADV SPI bus and multiplexes GPIO8/9 between the TCA8418 keyboard controller and NRF24 CE/CS.

The implementation:

• adds the dedicated Cardputer ADV 3-in-1 build environment;
• configures CC1101 GDO0=13 / CS=15;
• configures NRF24 CE=8 / CS=9;
• keeps inactive SPI devices deselected;
• arbitrates GPIO8/9 between the TCA8418 keyboard and NRF24;
• prevents generic I2C recovery from misinterpreting NRF24 CE LOW as a stuck SDA line;
• keeps the NRF24 spectrum handling changes scoped to this hardware configuration.

This PR is rebased onto dev (fe07bd7). The only conflict was src/modules/NRF24/nrf_spectrum.cpp, rewritten by #2758 (SpectrumPlot, continuous trace, frequency ruler). The 3-in-1 changes were reapplied on the new structure:

• scanChannels() — the mux critical section wraps the sweep; the tolerant RPD detection stays scoped to #if defined(CARDPUTER_ADV_3IN1_MUX), and the #else path is the current upstream scanner (NRF_FULL_SCALE);
• nrf_spectrum() — dev early-returns when nrf_start() fails, so adv3in1Nrf24CriticalEnd() runs before that return; otherwise the mutex would leak and leave the TCA8418 keyboard stuck when the radio fails to start.

nrf_common.cpp, bus_HAL.cpp and the board files rebased cleanly.

Types of Changes

New Feature / Hardware Support

Verification

Both environments build successfully:

• pio run -e m5stack-cardputer
• pio run -e m5stack-cardputer-adv-3in1

The standard Cardputer environment was also verified to retain its original compile-time pin configuration.

Hardware testing on M5Stack Cardputer ADV with the JosephCGS 3-in-1 expansion validated:

• Cardputer ADV boot
• microSD
• TCA8418 keyboard
• CC1101 receive
• LoRa
• NRF24 SPI initialization
• NRF24 spectrum reception
• keyboard operation after leaving NRF24 Spectrum

Testing

Both environments build successfully against dev (fe07bd7):

Standard environment:

• RAM: 130596 / 327680 bytes (39.9%)
• Flash: 4146136 / 8388608 bytes (49.4%)

3-in-1 environment:

• RAM: 130604 / 327680 bytes (39.9%)
• Flash: 4146276 / 8388608 bytes (49.4%)

The hardware checklist above was validated on the functional implementation prior to the rebase. The final rebased commit was build-tested on both environments against dev; a hardware retest is pending.

Linked Issues

None.

User-Facing Change

Add optional M5Stack Cardputer ADV support for the JosephCGS CC1101/NRF24/LoRa 3-in-1 expansion board.

[leandrofleury]

Compatibility scope:

• m5stack-cardputer → upstream behavior remains unchanged
• m5stack-cardputer-adv-3in1 → hardware-specific support for the JosephCGS 3-in-1 expansion

[leandrofleury]

Rebased onto dev (fe07bd7) and changed the PR base from main to dev.

The only conflict was in src/modules/NRF24/nrf_spectrum.cpp, which was rewritten by #2758 (SpectrumPlot, continuous trace, frequency ruler). The 3-in-1 changes were reapplied on top of the new structure:

scanChannels() — the mux critical section now wraps the sweep, and the tolerant RPD detection stays scoped to #if defined(CARDPUTER_ADV_3IN1_MUX). The #else path is the current upstream scanner, including NRF_FULL_SCALE.
nrf_spectrum() — dev early-returns when nrf_start() fails, so adv3in1Nrf24CriticalEnd() is called before that return; otherwise the mutex would leak and leave the TCA8418 keyboard stuck whenever the radio fails to come up.
Teardown wraps stopListening()/powerDown() and the CE LOW / CS HIGH sequence before freeing the plot buffers.

nrf_common.cpp, bus_HAL.cpp and the board files rebased cleanly.

Both environments build against dev.