Fix panic in scheduler due to potential timer overflow

[dbrgn]

Previously the code panicked when re-scheduling a measurement:

panicked at 'assertion failed: dur.inner < (1 << 15)', src/monotonic_stm32l0.rs:167:9

The reason is that we're scheduling the next measurement 5000ms into the future:

ctx.schedule
    .start_measurements(ctx.scheduled + 5000.millis())

Unfortunately our 16bit timer cannot do this:

Because the timer has 16 bits, it will overflow every 125 µs * 2^16 = ~8.19 seconds. Due to overflow checking, we can only safely use one half of the available timer range, meaning that we can safely schedule tasks 4.096 seconds into the future, with a resolution of 125 µs.

Thus, reduce the timeout to 4s. In the future, we'll use the RTC to schedule the next measurement.