Diagnose and skip rogue Nightscout profile records

[bjorkert]

Summary

• Switches the Nightscout profile fetch from /profile/current to /api/v1/profiles?count=1&find[startDate][$lte]=now, so future-dated records can no longer hide the active profile. /profile/current only sorts by startDate desc and ignores whether that date is in the past or future, which makes a single rogue upload (e.g. a phone with the wrong system clock) keep winning indefinitely. The find[] filter is the same query syntax already used for treatments, sage/iage/cage, BG data, and stats.
• Adds a "Run diagnostics" button in the Remote Settings → Debug section. It fetches 14 days of profile history and surfaces three known failure modes for remote commands:
  • Profile uploaded by a different app — when Loop and Trio share a Nightscout, profile uploads overwrite each other. Detected by checking which key path carries bundleIdentifier (Loop populates loopSettings.bundleIdentifier, Trio populates the top-level field) against Storage.shared.device.value.
  • Multiple devices uploading profiles — multiple installs (spare/old phones) each push their own deviceToken, causing APNS commands to fail intermittently. Detected by compressing consecutive same-token runs and warning when a token reappears (true alternation), not when tokens simply rotate (a→b→c stays silent).
  • Future-dated profile record found — informational only, since the new fetch already skips them; flagged because the rogue record still corrupts what Nightscout's own dashboard considers current and signals a broken phone clock somewhere.
• Results auto-scroll into view when the run completes (ScrollViewReader anchored to the diagnostics block).
• Reuses existing helpers — NightscoutUtils.executeRequest/parseDate, the find[] parameter convention, dateTimeUtils.formattedDate(from:), the orange exclamationmark.triangle warning idiom from ImportExportSettingsView. No new abstractions.

[bjorkert]

[bjorkert]