[Oura ring] - Promote Tier B (unverified) steps to step count

[pipiche38]

The problem

NOOP is receiving Oura's MET activity stream but refusing
to persist/display it because it's unverified and isn't actually a step count.

The idea

To make activity/steps visible we need to promote the data out of Tier B:

1. Capture real 0x50 (and ideally 0x7E/0x7F) fixtures and validate the decode against ground truth → reclassify from tierB to tierA in EventTags.swift.
2. Add a mapping case in OuraStreamMapping.swift (a stepSample row for real_steps, and/or an activeKcal/MET row for 0x50) — being honest about MET-vs-steps.
3. Get the realSteps server flag enabled so the ring actually sends step data.

[pipiche38]

Phase 1 — Instrument & log an estimate, persist nothing

• Add a pure OuraActivityEstimator (in OuraProtocol or StrandAnalytics, so it's unit-testable) that accumulates MET samples into a rolling window and produces: activeMinutes (MET ≥ threshold), estActiveKcal (MET × bodyweight × duration), and optionally a clearly-labelled MET-derived "step proxy".
• Replace the spammy per-record activity (Tier-B) state=… met=… log with a periodic aggregate log (every N minutes and on disconnect), e.g. Oura activity estimate [19:00–08:30]: samples=812 activeMin=41 estKcal=310.
• Unit tests against the PR #960 MET fixtures.
• validate over several days against Whoop (as I wear the 2) before anything is stored.

Phase 2 — Promote decode Tier-B → Tier-A

• Add real-capture fixtures, lock the decode byte-for-byte, flip the tag's tier in EventTags.swift, keep MET labelled as MET (no relabelling to steps).

Phase 3 — Persist & surface

• Add a mapping case in OuraStreamMapping.swift → a durable stream (new active-minutes/MET table, or reuse the activeKcalEst path).
• Aggregate into dailyMetric (AnalyticsEngine), then surface on the STEPS card (or a new ACTIVITY card).
• Mirror in the Android twin (android/) for parity.

Phase 4 — Day-owner + full-day totals

• Ensure the Oura device's activity feeds dailyMetric and DayOwnerResolver picks it; note that a full day total still depends on fixing the
  history-fetch cursor regression (live-only accumulation is enough for Phase 1 validation).

[pipiche38]

On-going work

• https://github.com/pipiche38/noop/tree/oura-integration-clean

[pipiche38]

@ryanbr feel free to assign this to me

[pipiche38]

Phase 1 implemented in #143

[pipiche38]

investigating and progressing on https://github.com/pipiche38/noop/tree/oura-real-steps-enable