support for usage

[pelikhan]

The server returns the token usage when asked nicely (with some json).

IMO, it would be very useful to inform the user about the costs of any LLM operation.

[marcelsafin]

Hey! I'm working on implementing this. Building a gh models usage command that calls the /users/{username}/settings/billing/premium_request/usage endpoint to show a per-model breakdown of premium request usage and costs.

Features planned:

• Per-model request counts and costs (gross/net)
• --today flag for daily usage
• --month / --year flags for historical queries
• Table output using the existing tableprinter pattern

Will open a PR shortly. Refs #81