feat: add --inference flag to token create command to create bearer tokens usable in the inference gateway

[shawnfeldman]

Summary

• Adds --inference flag to lk token create that sets an InferenceGrant{Perform: true} on the generated JWT
• Adds "Inference" option to the interactive permission multi-select prompt

Example

$ lk token create --inference --identity inference_e2e --valid-for 60m
Using generated room name [room-12314]
valid for (mins):  60
Token grants:
{
  "room": "room-12314"
}

Access token:....

Decoded JWT payload confirms the inference grant:

{
  "identity": "inference_...",
  "inference": {
    "perform": true
  },
  "name": "inference_..",
  "video": {
    "room": "room-12314"
  }
}

Test plan

• lk token create --inference --identity inference_e2e --valid-for 60m generates a valid JWT with inference grant
• lk token create interactive prompt shows "Inference" as a selectable permission
• Existing token creation flows (join, create, admin, etc.) are unaffected
• go test ./... passes

🤖 Generated with Claude Code

[CLAassistant]

All committers have signed the CLA.

[rektdeckard]

Nice! If you want to bump the version as well, I can kick off a release for this.

[rektdeckard]

This should probably print the whole ClaimGrants struct, not just the VideoGrant.