Small fixes: airtime fallback underflow, LLCC68 wrapper build, CLI gps crash, host test build

[mmmorks]

Staged on the fork for review. Final destination: meshcore-dev/MeshCore dev (or the fork; to decide). Base here is staging/meshcore-dev so the diff shows only this change.

Summary

Five small, independent fixes, each self-contained in its own commit hunk.

calcMaxPacketMillis() fallback underflow (src/helpers/radiolib/RadioLibWrappers.cpp). When getTimeOnAir() returns nothing usable (it returns 0 on an unconfigured modem) the fallback said "4 secs" but used 4000 - preamble_us, i.e. 4 ms minus a preamble that exceeds it at every setting there is. The unsigned subtraction underflows, and the result is not a mis-sized deadline but the absence of one: a ~49-day payload watchdog leaves CustomSX1262::isReceiving() holding a latched HEADER_VALID true forever, so the channel never again reads idle. The fallback is now a flat 4 s of payload (MAX_PACKET_FALLBACK_PAYLOAD_US, overridable) with no subtraction left to underflow.

CustomLLCC68Wrapper::doResetAGC() does not compile. It calls sx126xResetAGC() with one argument; the only overload takes (SX126x*, bool rx_boost_gain). Nothing in the tree currently includes this wrapper, which is why no build has failed, but the first board to use it will. Pass getRxBoostedGainMode() like the SX1262/SX1268/STM32WLx wrappers do.

gps CLI command crashes when no GPS was detected (src/helpers/CommonCLI.cpp). The status branch does strcmp(_sensors->getSettingByKey("gps"), "1"), and getSettingByKey() returns NULL when no gps setting is registered — which is the case whenever gps_detected was false at boot. Treat a missing setting as "deactivated".

Failed-receive diagnostics (RadioLibWrappers.cpp). The readData() error line now also prints the packet length, RSSI and SNR (quarter-dB, the same convention as Packet::_snr). The modem writes its packet-status registers whether or not the CRC passed, so this costs no extra SPI traffic and is enough to tell a failure distribution sitting on the SF's SNR floor apart from one spread across strong signals (a collision pattern).

Host test build. ConfigSerializer.cpp uses atoi/atol/atof and reaches them through Arduino.h on MCU targets, but not in the native googletest environment: on macOS the whole suite fails to compile at ConfigSerializer.o. Include <stdlib.h>. Likewise LocationProvider::sendSentence() was declared virtual but never defined; every current subclass overrides it, so MCU links get away with it, but any translation unit that emits the base vtable gets an undefined reference. Give it an empty default body, matching the no-op override in EnvironmentSensorManager.cpp.

How it was tested

• pio test -e native: all suites pass on macOS (before the <stdlib.h> fix none of them compile there).
• Compiled for heltec_tracker_v2_repeater (ESP32-S3) and WioTrackerL1_repeater (nRF52).

Dependencies

Independent. Applies to dev.