fix(rp2040): use int64 in ReadTemperature to avoid int32 overflow

[rdon-key]

Summary

This PR fixes the implementation of ReadTemperature for the RP2040 target.
The previous code used int32 intermediates, which caused arithmetic overflow when the ADC voltage deviated more than ±56 mV from 706 mV.
Since the internal temperature sensor of the RP2040 typically operates between 537 mV and 821 mV (-40°C to 125°C), overflow occurred in normal conditions.
As a result, the function sometimes returned corrupted values such as -31.8°C.

Changes

• Promote intermediate calculations in ReadTemperature (RP2040) to int64.
• Keep the return type int32 (milli-°C) for API compatibility.

Before (buggy)

temperature: 32.8℃
temperature: 32.8℃
temperature: -31.8℃
temperature: -31.8℃
...

After (fixed)

temperature: 31.8℃
temperature: 32.8℃
temperature: 32.8℃
temperature: 32.8℃
temperature: 33.7℃
...

Impact

• Prevents overflow in the normal RP2040 sensor range (-40°C to 125°C).
• Produces stable, monotonic readings without spurious negative jumps.
• No breaking API change.

[rdon-key]

Diff int32 to int64

v_hex	mV	orig_mC	orig_degC	ref_mC	ref_degC	diff	monoBreak
0x02B8127F	696.072250	-32768	-32.768	32768	32.768	true	false
0x02B81280	696.072266	-32768	-32.768	32768	32.768	true	false
0x02B81281	696.072281	-32768	-32.768	32768	32.768	true	false
0x02B81282	696.072296	32767	32.767	32767	32.767	false	true
0x02B81283	696.072311	32767	32.767	32767	32.767	false	false
0x02B81284	696.072327	32767	32.767	32767	32.767	false	false
0x02B81285	696.072342	32767	32.767	32767	32.767	false	false
0x02B81286	696.072357	32767	32.767	32767	32.767	false	false

[deadprogram]

Thanks for looking into this @rdon-key