fix(discord): align voice docs and slash UI

[EdderTalmor]

Closes #11349

Summary

• Document Discord proxy support and the Discord voice packet dump env var.
• Correct the Discord privileged-intent guidance so Server Members Intent is described as optional for numeric allowlists.
• Update the Discord voice docs to match the current SPEAKING-event SSRC mapping and barge-in behavior.
• Add /voice join to the slash-command autocomplete and cover it with tests.

Verification

• pytest tests/gateway/test_voice_command.py -q
• pytest tests/gateway/test_discord_opus.py -q

[alt-glitch]

Related to #11350 — appears to address the same 6 documentation drifts from #11349. May be a duplicate PR.

[teknium1]

Superseded by the salvage of #11350 (landed as #20411 on main), which applied the /voice join slash Choice and the correct Server Members Intent fix. The HERMES_DISCORD_VOICE_PACKET_DUMP env var proposed in this PR would be a new behavioral env var; per our policy, behavior switches belong in config.yaml rather than HERMES_* env vars — if packet dumping for voice debugging is worth keeping, please open a fresh PR that wires it through config.yaml (e.g. gateway.discord.voice.packet_dump). Thanks @EdderTalmor.