Last read milliseconds check skipped at 48-day rollover

**Describe the bug**
In I2c sensors there's a check on last red milliseconds = 0 or now minus last red is greater than the pole period, which could feasibly happen as the component is first added I'm a causing unnecessary duplicate read meaning in v1 different sensor metrics could be reported from different sampling periods but in V2 shouldn't t be an issue because all metrics share the same sampling period.

Start by locating the I2c sensor check that compares the last-read milliseconds with the current time and polling period. Reproduce or reason through the 48-day millisecond rollover and confirm whether it causes an unnecessary duplicate read. Done means the rollover case no longer skips or duplicates the intended sampling behavior, while preserving the stated V1 and V2 sampling expectations.

Written by the indexing model from the issue text.