utilities/gadgetbridge_to_influxdb#13: Time spent at stress levels

We currently capture stress levels over time (i.e. multiple point in time levels), but those don't make it particularly easy to chart out the proportion of time spent at each stress level.

In zepp_to_influxdb I was able to build a stream of points showing stress level - this was possible because each stress entry gave a start + end date.

It was then possible to iterate over those points with a Flux query to build a piechart

import "date"

from(bucket: "health")
  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)
  |> filter(fn: (r) => r["_measurement"] == "zepp")
  |> filter(fn: (r) => r["_field"] == "current_stress_level")
  |> filter(fn: (r) => r["stress"] == "point_in_time")

  // Exclude sleeping hours
  |> filter(fn: (r) => not contains(value: date.hour(t: r._time), 
                       set: [0,1,2,3,4,5,6]))
  // Skip missed reads
  |> filter(fn: (r) => r._value > 0)

  // Turn the reads into thresholds and counts
  // the thresholds used come from the description
  // in the apps UI
  |> map(fn: (r) => ({
    _time: r._time,
    _value: 1,
    _field: if r._value <= 39
               then "relaxed"
               else if r._value >= 40 and r._value <= 59
                    then "normal"
                    else if r._value >= 60 and r._value <= 79
                         then "medium"
                         else if r._value > 80
                              then "high"
                              else "unknown"

  }))
  // Sum the values so we can get the relative proportions
  |> sum()
  |> keep(columns: ["_field", "_value"])

I'd actually quite like to be able to cut Flux out of the loop on this, but even if that's not (immediately) possible, we need to find a way to generate 1 minute readings indicating stress level so that they can be consumed by something like that query.