fix(whatsapp): normalize bare phone targets to JIDs

[sgaofen]

Summary

• normalize bare WhatsApp phone numbers to @s.whatsapp.net JIDs before sending them to the local bridge
• apply the normalization in both the gateway adapter and the send_message tool path
• add regression tests covering bare-phone normalization for both direct bridge sends and adapter sends

Root Cause

Both tools/send_message_tool.py and gateway/platforms/whatsapp.py passed raw phone numbers directly to the Baileys bridge. Baileys expects a WhatsApp JID like 50766715226@s.whatsapp.net, so bare numbers caused jidDecode(...) to return undefined and the bridge crashed with HTTP 500.

Closes #8637.

Testing

• python3 -m py_compile gateway/platforms/whatsapp.py tools/send_message_tool.py tests/gateway/test_whatsapp_connect.py tests/tools/test_send_message_tool.py
• uv run --extra dev pytest tests/gateway/test_whatsapp_connect.py -q
• uv run --extra dev pytest tests/tools/test_send_message_tool.py -q -k 'whatsapp'

Platform Tested

• macOS 15.x (Apple Silicon)

Contribution Guide Notes

• Reviewed CONTRIBUTING.md and checked for existing open PRs before submitting this scoped bug fix.
• Ran the targeted verification commands listed above for this PR. I have not claimed a full repo-wide pytest tests/ -q pass unless explicitly noted.

[alt-glitch]

Superseded by #15652 which addresses the same bare-phone JID normalization at both gateway and bridge layers. Both close #8637.

[alt-glitch]

Superseded by #15652 which addresses the same bare-phone JID normalization at both gateway and bridge layers. Both close #8637.

[teknium1]

Merged via #50379. Your fix was cherry-picked onto current main with your authorship preserved in git history (commit a4b1554).

The adapter had been migrated to a bundled plugin and the WhatsApp identity helpers extracted since you opened this, so the normalizer now lives in gateway/whatsapp_identity.py as to_whatsapp_jid() (the outbound inverse of normalize_whatsapp_identifier) and is wired at all six outbound bridge sites. One tweak on top: the :device@domain legacy form now collapses correctly instead of producing a double-@. Thanks for the fix and the tests!