fix(adb): heal wedged offline transport + Player.log boot gate (#18)

[Jordan231111]

What & why

Fixes the #18 failure on a low-end 2013 HP Pavilion where the Android-11 guest boots fine (multi-minute boot, home screen visible) but the script loops on device offline forever and never roots.

Root cause: on a slow boot the per-instance adb transport is created during early boot and wedges in offline (TCP socket up, adb handshake never finalized). adb's connect no-ops on an existing transport (already connected), so it never re-handshakes — Boot-And-Wait polled getprop sys.boot_completed over a dead transport until timeout. The reporter's manual kill-server; connect (after the home screen was up) recovered it — which is exactly what this automates. On a fast PC the offline window is a few seconds and the socket churns/handshakes on its own, so the bug never reproduced.

Changes (tools/bsr_magisk.ps1 → re-embedded into blueStackRoot.cmd)

• Heal the transport. Per poll: connect → get-state; if not device, disconnect+connect to force a fresh handshake, then re-check. Deterministic recovery instead of relying on the socket happening to churn.
• Host-side boot gate. Player.log is per-instance tagged with a phase (StartingKernel → StartingAndroid → Ready). [Ready] is a "guest booted" signal independent of adb (immune to the offline race). Marked at launch so a previous boot's [Ready] can't false-positive.
• Liveness, not WMI-only. Relaunch fires only when nothing says the instance is alive (conf adb port not listening AND no Player.log activity AND no process), so the false Get-HdPlayerCount==0 (null/blocked WMI command line) no longer spams retrying launch.
• Sane timeouts. Fail fast (~90s) when nothing is alive at all (not a slow boot); cap the post-[Ready] wait at 120s; otherwise tolerate multi-minute boots.
• Retry helpers heal too. AdbShellRetry/AdbTry now disconnect+connect on a dropped transport instead of a no-op reconnect (covers mid-DATA drops such as the magiskd reboot — the reporter's second-boot wedge).

Logging split (intentional)

• blueStackRoot.cmd (production / all future releases): terminal logging only — the existing Say/Write-Host progress output, no files written.
• debug.cmd (standalone read-only probe): the foolproof logger — launches the instance, logs get-state / Player.log / WMI, runs the disconnect+connect heal live, and writes a full bsr_debug_*.log to the Desktop with a final VERDICT (SUCCESS via HEAL, booted-but-adb-stuck, never-started, crashed). Designed so one attached log is enough to diagnose.

Tests

• New unit tests for Parse-AdbState and the Player.log ready/alive parsers (via a PlayerLogProbe seam).
• Run-Magisk-Unit-Tests.ps1: 256 passed, 0 failed.
• reembed.ps1 + Check-Embedded-Sync.ps1: embedded blueStackRoot.cmd blocks in sync with tools/ sources; the production .cmd writes no log files (bsr_run_/Init-BsrLog/LogV = 0).

Refs #18