[ryanbr]

Follow-up to #1579, which I flagged when reviewing it and recorded in that merge commit.

#1579 stopped a partial sample-day count activating the 4.0-only calibration affordance on a WHOOP
5.0. The coefficient paths could still activate it, and those are profile-global — so a user who
calibrated a 4.0 and later moved to a 5.0 kept getting the 4.0 prompt on any day the 5.0 logged no
steps and no estimate existed. That is the complaint issue #1523 opened, on a narrower trigger.

Why the original justification doesn't hold

@kavemang kept those paths to preserve estimate behaviour across that migration. Reasonable on its
face — but this gate doesn't control the estimate:

• estSteps comes from stepsEstByDay[selectedDayKey] (TodayView.swift:3737), computed
  independently;
• the gate is only consulted inside
  needsCalibration = realSteps == nil && estSteps == nil && stepsPipelineActive(…).

So all it decides is whether a blank tile offers to calibrate. A strap that reports steps natively
has nothing to calibrate.

The change

A positively-identified 5/MG returns false before anything else is considered.

Android has been immune this way all along — stepsCalibrationPrompt returns early on
model != WhoopModel.WHOOP4.name before reading any calibration state. This brings iOS to the same
guarantee while keeping its more permissive fallback for unknown straps.

What is deliberately NOT changed

The coefficient paths are not deleted, which is the tempting reading of the above. A legacy 4.0
owner whose selectedWhoopModel key was never written has no family to match on, and only the
coefficient says they are mid-estimate. Dropping them would silently take the calibration gear away
from exactly the users #1491 restored it for. A new case pins that:

legacy 4.0, pref unset, has coefficient  ->  still active

Matrix, replayed before changing anything

case	#1579	here
5.0 + partial sample days	false	false
4.0 fresh + day data	true	true
4.0, no day data	false	false
5.0 + fitted coefficient	true	false
5.0 + manual coefficient	true	false
unset model + sample days	false	false
legacy 4.0, pref unset, coefficient	true	true
4.0 calibrated + day data	true	true

Only the two intended cases move. Those two tests from #1579 are renamed to say what they now assert,
with the reasoning recorded on them rather than buried in a commit message.

Verification

• doc_comment_lint.py OK. No strings, no Android change, no schema.
• app-build required — TodayView.swift is app-target Swift and nothing else compiles it; that is
  how Fix WHOOP 5 steps calibration prompt (#1523) #1579's own tests went unrun until merge. Not dispatched.
• Not seen on a device.

Credit to @kavemang: the diagnosis and the pure-function extraction in #1579 are what made this
testable at all — the gate was previously unreachable from any test.