the day and the bout disagree about what counts as active

[abdulsaheel]

Calories.dailyEnergy gates active energy at a fixed fraction of HRmax
(activeFraction = 0.50). Calories.estimateBoutCalories gates at
restingHr + 0.30 * HRR. They are different quantities and they disagree, so
the same minute of the same day is billed active by one and resting by the
other:

profile	HRmax	day gate	bout gate	apart
70 y, RHR 68	159.0	79.5	95.3	+15.8
34 y, RHR 55	184.2	92.1	93.8	+1.7
25 y, RHR 45	190.5	95.2	88.7	-6.6

Note the sign flips. It is not a constant offset one path could absorb.

The fixed fraction runs the wrong way

0.50 * Tanaka is 104 - 0.35*age, so it falls with age while resting HR
generally does not. Measured against rest rather than against HRmax:

• 70 y with RHR 68: the gate sits 11.5 bpm above rest. Standing up,
  digesting a meal, a stressful phone call all clear it, so the day accrues
  active surplus from existing.
• 25 y with RHR 45: the gate sits 50.2 bpm above rest. Real walking at
  HR 90 bills zero active energy.

So it over-counts the old and deconditioned and under-counts the young and fit.

Published HR-FLEX (Spurr 1988; Ceesay 1989) determines FLEX individually, as
the midpoint between the highest resting HR and the lowest HR during light
imposed exercise. It is individual precisely because a fixed threshold does not
transfer between people.

Keytel is at its most non-linear right at that gate

Male, 34 y, 72 kg at HR 92 (about the flex point):

-55.0969 + 0.6309*92 + 0.1988*72 + 0.2017*34 = 24.1173 kJ/min
                                             =  5.764 kcal/min  (~4.6 METs)

HR 92 in a 34-year-old is light walking, roughly 3 METs. The model over-reads by
about 60% exactly where the gate opens. Mifflin basal is 1.158 kcal/min, so
~4.6 kcal/min of "active surplus" is credited for every minute of
near-sedentary time that clears the threshold. On the older profile above,
where the gate is 11.5 bpm over rest, that is a lot of minutes.

This is a larger error source than anything else currently in the calorie path.

Proposal

One gate, HRR-relative, shared by both entry points — the bout form is the
better of the two, and it at least scales with the individual's own resting HR.
Then look separately at whether the low-HR end of Keytel should be blended
toward the resting rate rather than switching at a hard threshold, since the
discontinuity at the gate is where the over-read concentrates.

Changing this moves persisted numbers, so it needs a kAlgoVersion bump on the
edge side and a note that historical days will shift.