### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### The Feature

The Fireworks AI serverless catalog now includes a `glm-5p2-fast` router endpoint (`accounts/fireworks/routers/glm-5p2-fast`). This router has no entry in `model_prices_and_context_window.json`, so cost attribution and capability detection are missing for callers using it through LiteLLM.

Adding the router entry (plus a short-form alias `fireworks_ai/glm-5p2-fast`) would bring the registry in sync with the current Fireworks platform catalog. Per-token costs, context/output limits, `cache_read_input_token_cost`, and capability flags come from the Fireworks model page at https://app.fireworks.ai/models/fireworks/glm-5p2.

This also corrects the pre-existing `glm-5p2` direct model entry: `max_input_tokens` was 1048576 but should be 1048575, and `cache_read_input_token_cost` was 2.6e-07 but has been lowered to 1.4e-07 ($0.14/M) per the current Fireworks pricing. The same context window correction applies to the new `glm-5p2-fast` entry.

### Motivation, pitch

Accurate spend tracking and correct token-limit enforcement are core proxy features. Callin…