Feat/airspy stream owner diagnostics - #687

The conventional scanner was opening and immediately cancelling a new
StreamIQ session on every channel hop (~100 ms dwell). The Airspy driver
needs a brief teardown window before a new transfer can start, so rapid
close/reopen produced 'airspy: stream already active' on every other
channel advance.
Fix: open one IQ stream at Run() startup and keep it alive for the
lifetime of the scanner. SetCenterFreq retunes the hardware; the
running stream delivers samples at the new frequency within a few ms.
The stream is only reopened if the device disconnects (stream channel
closes), providing natural recovery without close/open churn.
Internal changes:
- Scanner.Run: open one persistent stream; reopen only on stream-died
- Scanner.scanWindow: return (broken, streamDied bool) so Run can
 distinguish device disconnect from normal dwell expiry
- Scanner.beginDwell: take ctx instead of streamCtx/cancel; return
 streamDied bool; no longer cancels/owns the stream lifetime
Broker-side diagnostics added in this branch:
- cmd/gophertrunk/daemon.go: iqPrimaryOwner map + owner labels on all
 openSingleChannelIQ call sites and on cc/wideband primary claims
- diagnosticIQSource wrapper on the conventional scanner IQ path logs
 every direct StreamIQ open attempt/result at DEBUG level (WARN on
 failure) so stream contention is identifiable from logs with
 log.level: debug
Docs: install-windows.md troubleshooting table now covers
'conv: StreamIQ failed / airspy: stream already active' with guidance
on reading the new diagnostic log fields (primary_owner, owner).
Validated on host DBI-NUC-01 (Airspy R2, Windows 10.0 build 26300):
- Before: continuous WARN 'conv: StreamIQ failed' loop at ~3 Hz
- After: single 'conv: opening IQ stream' at startup; 18 min stable
 runtime with zero StreamIQ errors; goroutine count stable
 at 36-37; heap steady at 2-3 MB

...stics