ConfigHWCompass: show calibration bad-fit reasons in mag cal dialog

[christianpetri]

Summary

Surface compass calibration failure reasons (Mag(N) bad ... STATUSTEXT) directly inside the onboard mag calibration dialog, so users can see why a calibration was rejected without leaving the page.

Pairs with the ArduPilot side that emits these diagnostics: ArduPilot/ardupilot#32757.

What changed

In both ConfigHWCompass (legacy) and ConfigHWCompass2:

• Subscribe to STATUSTEXT while calibration is running (in addition to the existing MAG_CAL_PROGRESS and MAG_CAL_REPORT subscriptions).
• Filter messages with the regex Mag\((\d+)\) bad and store the latest message per compass ID in a thread-safe Dictionary<string,string>.
• Render the latest bad-fit message per compass under the progress and report lines in lbl_obmagresult.
• Auto-scroll the textbox so the newest entries are visible.
• Switch line breaks to Environment.NewLine so newlines render correctly in the WinForms TextBox.
• Return false from the new STATUSTEXT branch so non-compass STATUSTEXT continues to flow to other subscribers (HUD/messages remain unaffected).
• Clear collected messages on Start; unsubscribe on Accept and Cancel.

Why

Previously a failed compass calibration just looped silently with no indication of the cause. With ArduPilot now emitting specific Mag(N) bad ... reasons (bad fit, bad radius, bad orientation, etc.), Mission Planner can show these inline so the user immediately knows what to fix.

Tests

New MissionPlannerTests/GCSViews/ConfigHWCompassTests.cs (MSTest):

• Regex matches each known bad-* variant and captures the compass ID.
• Regex rejects unrelated STATUSTEXT (PreArm, EKF, etc.).
• Latest-per-compass dictionary semantics validated (replacement, not accumulation).
• ASCII NUL-trim behavior validated.

Result: 7/7 passing.

Manual verification (SITL)

• Successful calibration: progress + report lines shown, no bad-fit lines.
• Forced failure (tight COMPASS_CAL_FIT, extreme SIM_MAG*_OFS_*): Mag(N) bad ... lines appear in the dialog, replaced on each retry rather than accumulating.
• Multi-compass failure: one current bad line per compass ID.
• Cancel + restart: state clears cleanly.
• Normal HUD/messages stream still receives non-compass STATUSTEXT.

Notes

• No new parameters, no MAVLink changes.
• Behavior unchanged for successful calibrations.
• AI-assisted contribution.

[christianpetri]

Closing this in light of where ArduPilot PR #32757 ended up. The original approach here captured `Mag(N) bad ...` STATUSTEXT messages emitted from `fit_acceptable()`, but that GCS_SEND_TEXT approach was rejected during review due to flash cost (~448 bytes). The final version of #32757 emits no STATUSTEXT — it only sets new enum values (BAD_OFFSETS=8, BAD_DIAG=9, BAD_FITNESS=10) in `MAG_CAL_REPORT.cal_status`. That makes the STATUSTEXT subscription, regex, and dictionary added here dead code that will never fire. The right fix for issue #3709 is a much smaller change: add the three new values to the `MAG_CAL_STATUS` enum in `ExtLibs/Mavlink/Mavlink.cs`. The existing `(MAVLink.MAG_CAL_STATUS)obj.cal_status` cast in both compass forms will then display named labels automatically, with no extra subscription or logic needed. Will open a replacement PR once the upstream mavlink/mavlink XML is merged.