Align log tags with actual class names:
- [CSASession] -> [eDVBCSASession]
- [CSAEngine] -> [eDVBCSAEngine]
- [SoftDecoder] -> [eDVBSoftDecoder]

When a CW arrives from OSCam before ECM analysis completes and the
CSA session activates, the CW is buffered (PendingCw) and replayed
immediately upon activation. This handles the race condition where the
CW arrives ~50-250ms before the session is ready.

When switching from StreamRelay to Live-TV for the same service, the
softcam's demux is already running and a simple LIST_UPDATE causes it
to just "continue processing" without resending the current CW. The
new CSA session has no keys and must wait for the next ECM cycle (~4-8s).

Fix: Send CMD_NOT_SELECTED first to stop the demux, then LIST_ADD with
CMD_OK_DESCRAMBLING to restart it. This forces the softcam to treat the
service as new and immediately send the current CW.

…ryption

When a service uses CSA-ALT and the ECM cache has a hit, the CSA session
activates speculatively via startECMMonitor() during eventNewProgramInfo.
However, recheckPMTHandlers() runs AFTER this event and may assign a CI
slot for hardware descrambling. This caused the SoftCSA icon to show
incorrectly and the SoftDecoder to run unnecessarily alongside the CI module.

Add eventCIConnected to eDVBServicePMTHandler, fired after recheckPMTHandlers()
assigns a CI slot. eDVBServicePlay handles this event by stopping the ECM
monitor and deactivating the CSA session, which triggers onSessionActivated(false)
to stop the SoftDecoder and restore the hardware decoder.

distributeCAPMT() was sending CAPMT objects for services that had been
registered but never received their PMT. Since m_capmt[2048] was not
initialized in the eDVBCAService constructor, this resulted in stale
heap data (typically recycled EIT/EPG section data) being sent over
the socket, corrupting the DVBAPI protocol stream.

This caused "Unknown socket command received" and "network packet
malformed" errors in OSCam when zapping between channels (particularly
BISS channels on Abertis 30W where active EPG data fills the heap).

Fix: Skip services with m_version == -1 (buildCAPMT never called) in
distributeCAPMT(), and zero-initialize m_capmt in the constructor.

…oder

When switching between SoftCSA channels, the skin was notified via
evUpdatedInfo immediately after SoftDecoder takeover. At that point the
SoftDecoder's hardware decoder does not exist yet (it is created only
after the first CW arrives), so getVideoHeight() returns -1. The skin
falls back to eAVControl which reads /proc/stb/vmpeg/0/yres from the
just-released HW decoder, returning 0. This causes the HD icon to grey
out. On platforms where VIDEO_EVENT_SIZE_CHANGED is suppressed for
identical resolutions (e.g. VU Solo4K), the icon never recovers.

Add a m_decoder_ready signal to eDVBSoftDecoder, fired after the decoder
reaches PLAY state. eDVBServicePlay defers the evUpdatedInfo notification
until this signal arrives, ensuring the skin queries valid video info.