LKMA-216 add V.22bis softmodem path

[nicolasumaras]

Summary\n- advertise V.21 plus V.22/V.22bis in the SpanDSP V.8 answer menu and select V.22bis when offered\n- add V.22bis data-mode diagnostics plus HDLC pty bridge coverage for the worker path\n- document the real Windows lab deployment/runbook flow for IPCP-open and ppp0 ping validation\n\n## Verification\n- npm test\n- make -C vendor/sipfax-softmodem SPANDSP_CFLAGS='-I/tmp/tmp.fur9ICi9J7/root/usr/include/x86_64-linux-gnu -I/tmp/tmp.fur9ICi9J7/root/usr/include' LDLIBS='-L/tmp/tmp.fur9ICi9J7/root/usr/lib/x86_64-linux-gnu -Wl,-rpath,/tmp/tmp.fur9ICi9J7/root/usr/lib/x86_64-linux-gnu -lspandsp -lm'\n- LD_LIBRARY_PATH=/tmp/tmp.fur9ICi9J7/root/usr/lib/x86_64-linux-gnu make -C vendor/sipfax-softmodem bench-v21 bench-v22bis bench-hdlc-pty\n

[nicolasumaras]

Approve — LKMA-216 Phase 5a V.22bis path

Reviewed PR head 862702c601da73445172f3aa8c7c0ece00f31190 on branch lkma-216-v22bis-path (task branch off main, not stacked on lkma-199-pppd-egress).

What I checked:

• V.8 menu advertises V.21 + V.22bis — vendor/sipfax-softmodem/sipfax-softmodem.c:702 sets parms.modulations = V8_MOD_V21 | V8_MOD_V22.
• V.22bis selected on V.8 offer — v8_result at sipfax-softmodem.c:650-662 branches on V8_MOD_V22 into enter_data_mode(MODULATION_V22BIS, "v8-v22bis-selected"); falls back to V.21 on non-V.8 / failed V.8.
• V.22bis demodulator/modulator init — v22bis_init(... 2400 ...) at sipfax-softmodem.c:618-633 with HDLC pty bridge plumbed through hdlc_tx_next_bit and put_hdlc_bit.
• media.modem modulation diagnostics — Worker emit_control writes modulation/baud (sipfax-softmodem.c:325-353); Node side src/media.js:534-705 merges control events into lastControl and surfaces modulation, baud, state, ber, lastEvent, lastEventAt from diagnostics().
• V.22bis bench coverage — vendor/sipfax-softmodem/bench-v22bis.py plus make bench-v22bis proves V.22bis data-mode entry against artifacts/lkma-193a/ground-truth/v22bis/inbound.wav. README is explicit that the artifact is synthetic and the bench forces the modulation; acceptable for the "data-mode entry against the recording/replay artifact" deliverable.
• HDLC pty bridge coverage — test/hdlc-pty-loopback.py + make bench-hdlc-pty.
• Runbook docs — deploy/README.md "Real Windows Lab Run" (lines 219-314) lists prerequisites, wiring, expected media.modem.modulation: V.22bis / baud: 2400, IPCP-open confirmation, ppp.interfaceName check, and ping validation over ppp0.
• Proxmox lab access doc — deploy/README.md:27-33 documents the access path through root@192.168.1.20 and shared PROXMOX runtime secret without storing the secret value.

CI:

• Workflow CI job test: SUCCESS at head 862702c (run 28258403552).
• Mergeable: MERGEABLE / CLEAN.
• No prior reviews on the PR.

Deployment evidence (per Developer comment on LKMA-216):

• /opt/sipfax on VM 133 / 192.168.1.31 checked out to 862702c.
• Worker rebuilt and installed to /usr/local/bin/sipfax-softmodem.
• sipfax.service active; /healthz returns status: ok.
• Listeners confirmed: UDP 5060, UDP 40000, TCP 127.0.0.1:8080.
• Independent reachability probe: UDP 5060 and SSH 22 open from outside; operator 8080 correctly loopback-only.

Non-blocking follow-up tracked separately:

• Add a real V.8-negotiated V.22bis bench once a non-synthetic V.22bis recording is available; mirror the V.21 control-event test for V.22bis on the Node side.

Merging now per the LKMA-216 acceptance path.