Add one wire i2c bridge support (#10078)

* First version of DS248X bridge

* Add first iteration of DS248X sensor

* Supports single readings on DS2484
* Supports readings on ch0 for DS2484_800
* Detection of variant for DS248X

* Minor fix on retries for sensor init

* Allow multiple channel detect passes on 8-ch version

* Always read temperature via ROM matching

* Small comment to show how to send all channels

* Minor logging changes

* Prevent one-wire double definitions

* Detect ROMs per round

* Fix comment

* Prevent skipping on DS2482 ALT3 check

* Fix comment (again)

* Fix style  checks

* Remove comment for multiple measurements

* Address CodeRabbit review findings on DS248X sensor

Set _variant on every detectVariant path and branch on the member, so a
failed variant probe retries instead of falling into single-channel init.

Search DS2482-800 channels into a scratch buffer so a transient one-wire
failure cannot erase a ROM found on an earlier pass, and count channels
that already hold a ROM.

Check every one-wire return value in readTemperatureROM, validate the
scratchpad CRC, and return DS248X_INVALID_TEMPERATURE on failure so the
existing sentinel checks reject failed reads instead of reporting stale
or uninitialised data.

* Probe IIS2MDCTR WHO_AM_I before the DS248X status check

At HMC5883L_ADDR the DS248X probe reads 0xF0. On the IIS2MDCTR that
sub-address sets auto-increment and targets 0x70, which is reserved and
returns an unspecified value; any of bits 0x02, 0x04 or 0x10 makes the
probe claim the magnetometer as a DS2482.

Reading the WHO_AM_I at 0x4F first is deterministic for both parts. A
DS2482 does not acknowledge 0x4F, an invalid command code, and leaves
its read pointer untouched, so the subsequent read returns Status,
Configuration, Channel Selection or Read Data. None of those can hold
0x40 at scan time, so the DS248X probe still runs and detects it.

This also restores develop's detection order and matches the structure
already used at BMA423_ADDR: specific ID match, then probe, then the
generic fallback.

* Read only the reported channel in getMetrics

getMetrics walked all eight DS2482-800 channels, but only channel 0 is
ever written into the measurement. Each populated channel costs a
blocking 750ms conversion inside readTemperatureROM, so a fully wired
bridge stalled telemetry for roughly 6s per cycle and discarded seven of
the eight readings.

Channels without a sensor were already cheap thanks to the isValidROM
guard, so this only affects boards that actually use more than one
channel, which is the reason to fit a DS2482-800 in the first place.

Multi-channel reporting is handled separately in #10192; a note records
that it should start the conversion on every channel before waiting,
rather than reading each channel end to end.

* Trim DS248X comments to one line each

---------

Co-authored-by: Thomas Göttgens <tgoettgens@gmail.com>

src/modules/Telemetry/EnvironmentTelemetry.cpp

@@ -131,6 +131,10 @@ extern void drawCommonHeader(OLEDDisplay *display, int16_t x, int16_t y, const c
 #include "Sensor/BH1750Sensor.h"
 #endif
 
+#if __has_include(<Adafruit_DS248x.h>)
+#include "Sensor/DS248XSensor.h"
+#endif
+
 #define FAILED_STATE_SENSOR_READ_MULTIPLIER 10
 #define DISPLAY_RECEIVEID_MEASUREMENTS_ON_SCREEN true
 
@@ -239,6 +243,10 @@ void EnvironmentTelemetryModule::i2cScanFinished(ScanI2C *i2cScanner)
     // TODO Can we scan for multiple sensors connected on the same bus?
     addSensor<SHTXXSensor>(i2cScanner, ScanI2C::DeviceType::SHTXX);
 #endif
+#if __has_include(<Adafruit_DS248x.h>)
+    addSensor<DS248XSensor>(i2cScanner, ScanI2C::DeviceType::DS248X);
+#endif
+
 #endif
 }